Tag: agentforce

TL;DR: Some in the Salesforce ecosystem are wondering, “Is MuleSoft playing hide and seek?” as the branding takes a backseat to the Agentforce hype. The reality is that MuleSoft hasn’t disappeared; it has gone under the hood – evolved into the “Agent Fabric” – the essential infrastructure that allows autonomous AI agents to actually execute tasks. While Agentforce and Reasoning Engine, Atlas, provide the “brain,” MuleSoft provides the “limbs,” connecting AI to legacy data and external systems via governed APIs. Without this integration layer, an agentic enterprise is just a collection of smart chatbots that cannot actually move data or trigger real-world business actions.

We saw the buzz surrounding MuleSoft seemingly reach a fever pitch just before the Agentforce storm dominated the Salesforce ecosystem. To the casual observer, it might have appeared that promo of the integration giant was being sidelined by the new era of autonomous AI agents. However, the reality is the opposite. MuleSoft has not been replaced; it has been repositioned as the central nervous system of the “agentic enterprise”. Without the connectivity and governance MuleSoft provides to established tech stacks, Agentforce might be little more than a brain without limbs, capable of thinking but unable to act upon the world.

The Evolution from Integration to Orchestration

For years, MuleSoft was defined by its API-led connectivity. It was the tool used to bridge the gap between legacy systems like SAP or Oracle and modern cloud platforms. An organisation might use Salesforce for sales and service, Xero for accounting, Mailchimp for email marketing, Shopify for transactions, Shipmate for parcel tracking – they need a way for these systems to interact and speak the same language. Enter MuleSoft.

When Salesforce introduced Agentforce – a suite of autonomous AI agents capable of handling tasks across all business functions – the focus shifted from “how do we move data?” to “how does an agent take action?”.

For an AI agent to be truly useful, it must do more than answer questions; it must execute business processes. It needs to check inventory, process a refund, or update a shipping schedule in an external warehouse management system (WMS). MuleSoft provides the “Agent Fabric” – a governed layer of APIs that allows these agents to “reach out” of the Salesforce platform and interact with any other system in the enterprise.

Why MuleSoft is Essential for Agentic Enterprises

An agentic enterprise is an organisation where autonomous agents handle complex, multi-step workflows with minimal human intervention. This requires three things that MuleSoft is uniquely positioned to deliver:

• Unified Context: Agents need a single, real-time view of the customer. MuleSoft federates data from disparate silos and complex orgs – billing, ERP, and legacy databases – ensuring the agent has the full story before it makes a decision.
• Actionable Tooling: Through the Model Context Protocol (MCP), MuleSoft allows developers to turn existing integrations into “tools” that agents can discover and use. An API that was originally built for a mobile app can now be used by an AI agent to verify a warranty or reset a password.
• Governance and Guardrails: Giving an AI the power to execute transactions carries risk. MuleSoft’s API Manager provides additional security layers, such as rate limiting and PII detection, ensuring that agents operate within strict corporate policies.

Industry Use Cases: Stories of the Agentic Shift

The Retailer and the Supply Chain Crisis

Consider a fictionalised Fortune 500 retailer. Before the rise of agentic AI, their customer service team was overwhelmed during peak seasons. When a shipment was delayed, customers called in, and agents had to manually check three different systems: the order management system (OMS), the logistics partner’s portal, and the warehouse database.

By implementing Agentforce fed by MuleSoft, the retailer created a “Logistics Agent”. When a customer asks about a delay, the agent uses a MuleSoft-powered API to fetch real-time data from the carrier’s system. If the item is stuck, the agent doesn’t just report the news; it uses another MuleSoft “action” to trigger a proactive discount in the billing system and schedule a new delivery in the WMS. The integration layer turned a passive information-fetcher into an active problem-solver.

The Financial Institution and Open Banking

A leading European bank faced the challenge of “agentic fraud”. As they opened their systems to third-party providers via Open Banking, the volume of automated transactions spiked. They used MuleSoft to build a “Fraud Oversight Agent”.

This agent monitors API traffic in real time. When it detects an unusual pattern – perhaps an agentic script attempting to move funds across multiple accounts – it doesn’t just flag it. It uses MuleSoft’s orchestration capabilities to instantly freeze the specific API key and initiate a “Step-up Authentication” protocol, protecting the bank’s assets without human intervention.

Different Organisations, Unique Challenges

So, in the ‘Agentic Era’, what types of businesses are benefitting from MuleSoft?

Organisation Type

Primary Challenge

MuleSoft/Agentforce Benefit

Legacy-Heavy (e.g. Manufacturing)

Data trapped in on-premise systems with no modern interface.

MuleSoft creates “System APIs” that wrap legacy data in a format that AI agents can understand.

Highly Regulated (e.g. Healthcare)

Strict compliance (GDPR, HIPAA) prevents AI from accessing sensitive data.

MuleSoft’s governance layer masks PII before it ever reaches the AI model’s “brain”.

Fast-Growing Scale-ups

Brick wall silos caused by adopting too many tools. Applications built up over time in legacy systems or implemented too quickly in fast-growing orgs.

MuleSoft acts as a universal adapter, allowing new agents to be deployed across the entire tech stack in days.

The “Agentforce storm” did not blow MuleSoft away; it cleared the air to show exactly why a robust integration strategy is the only way to scale AI for organisations utilising many different platforms. As we move further into 2026, the enterprises that succeed will not be those with the smartest AI models, but those with the most “connected” ones. MuleSoft is no longer just about connecting apps; it is about empowering the next generation of digital workers to act with the full weight and intelligence of the entire enterprise behind them.

Don’t let your AI get stranded in a data silo. Bridge the gap between intelligence and action by making MuleSoft the backbone of your Agentforce strategy. Reach out today to turn your APIs into agentic tools.

References

Ksolves. (2026). Future of Mulesoft: Integration Trends to Watch in 2026. [online] Available at: https://www.ksolves.com/blog/salesforce/mulesoft-integration-trends [Accessed 15 Apr. 2026].

Polyakov, G. (2025). From APIs to AI: The New Integration Frontier. Psyncopate Technologies. [online] Available at: https://www.psyncopate.com/insight/ai-use-cases-with-mulesoft [Accessed 15 Apr. 2026].

RSM Technology. (2025). Building the Agentic Enterprise with MuleSoft: Dreamforce 2025 Recap. [online] Available at: https://technologyblog.rsmus.com/technologies/salesforce/building-the-agentic-enterprise-with-mulesoft-dreamforce-2025-recap/ [Accessed 15 Apr. 2026].

Salesforce. (2025). Architecting the Agentic Enterprise with MuleSoft. Salesforce Architects. [online] Available at: https://architect.salesforce.com/docs/architect/fundamentals/guide/mulesoft-architecting-agentic-enterprise [Accessed 15 Apr. 2026].

Scorch Agency. (2025). MuleSoft for Agentforce: Unleashing Greater Productivity. [online] Available at: https://www.scorchagency.com/wp-content/uploads/MuleSoft-For-Agentforce.pdf [Accessed 15 Apr. 2026].

Hazards businesses might miss before it’s too late.

TL;DR The meteoric rise of generative AI has often been described as a “gold rush,” but for many modern enterprises, AI is a “black ice” risk – transparent, hard to detect, and capable of sending a company into a reputational skid before they realise they’ve lost traction. While tools like Salesforce’s Einstein Trust Layer offer “traction control” by masking data and enforcing zero-retention policies, technical safeguards are not a cure-all.

The danger lies in Shadow AI, where employees bypass secure systems to use public, unsanctioned LLMs, inadvertently feeding proprietary data into public training sets. Combined with the staggering environmental costs (huge water and energy consumption) and a 95% failure rate for unguided AI pilots, the financial stakes are massive. To stay on the road, businesses must pair secure software with rigorous employee training and a commitment to Responsible AI.

***

In the physical world, verglas, also widely referred to as black ice, is a thin, transparent coating of glazed ice on a surface. It is notoriously difficult to see, often appearing as nothing more than a harmless wet patch until a vehicle’s tires lose grip. By the time the driver realises the danger, the car is already spinning out of control. For businesses in 2026, AI presents a comparable hazard. It promises a smooth, fast journey toward productivity, but beneath that polished surface lie invisible patches of “Shadow AI,” data leakage, and environmental costs that can send an entire organisation into a financial and reputational skid.

The Invisible Threat: Shadow AI

The most dangerous patch of black ice is Shadow AI – the unsanctioned use of AI tools by employees without the knowledge or oversight of the IT department.

While leadership may be debating official AI strategies in the boardroom, the workforce has already moved ahead. According to a 2025 KPMG global survey, up to 58% of employees use AI productivity tools daily, yet only 41% of organisations have a formal policy guiding that use (Espria, 2025). A concerning and undeniable governance gap.

When an analyst uploads a sensitive quarterly earnings report to a free, public Large Language Model (LLM) to “summarise the key trends,” they aren’t just saving time – they are potentially feeding proprietary corporate data into a public training pool. Without a protected license or “Enterprise” tier agreement, many LLM providers reserve the right to use prompt data to train future iterations of their models. Once that data is ingested, it is effectively public; it can reappear in the outputs of a competitor’s query months later. In this scenario, the business has lost its “grip” on its most valuable asset – its data.

Good News for Salesforce Users

For organisations already operating within the Salesforce ecosystem, there is a significant safety measure: the Einstein Trust Layer. The good news for Salesforce users is that this built-in governance framework acts as a sophisticated buffer between corporate data and generative models. By utilising dynamic data masking, the Trust Layer strips away personally identifiable information (PII) and sensitive records before they ever reach an external LLM, replacing them with anonymised placeholders.

Furthermore, Salesforce’s “zero-retention” policy ensures that no data shared through the platform is stored or used by third-party providers to train their public models. However, this technical safety net only extends as far as the Salesforce platform. The persistent danger remains that employees, if not rigorously trained on AI best practices, may still copy-paste that same sensitive data into consumer-grade LLMs or unsanctioned browser extensions outside of the Salesforce environment. Without a culture of “AI Literacy,” the security of the Trust Layer can be easily bypassed by a single well-intentioned but uninformed employee looking for a shortcut.

The Environmental and Financial Toll

The “black ice” metaphor extends beyond data privacy to the hidden costs of AI infrastructure. At high-level tech events, the conversation often centres on “efficiency” and “innovation,” but the physical reality of AI is resource-heavy and carbon-intensive.

A single exchange with an LLM can consume roughly 0.26 mL of water for cooling (Online Learning Consortium, 2025). While this seems negligible, at the scale of billions of monthly queries, the impact is staggering. Microsoft and Google reported year-on-year water consumption increases of 34% and 20% respectively as they expanded their AI-ready data centers (GOV.UK, 2025). AI servers are also projected to triple their energy demand by 2028, potentially consuming enough electricity to power 28 million homes (GOV.UK, 2025).

For businesses, these aren’t just ethical concerns; they are financial ones. Governments are increasingly mandating transparent environmental reporting. Companies that ignore the “green” cost of their AI implementations today may find themselves hitting a wall of carbon taxes and regulatory fines tomorrow.

Furthermore, the financial “spin-out” from failed AI projects is already a reality. Recent data indicates that 70% to 95% of all AI pilots fail to reach full production (Medium, 2026). These failures aren’t just lost time – they represent billions in “economic vandalism.” In 2025 alone, American companies spent an estimated $644 billion on AI deployments, many of which were abandoned due to poor data quality or escalating “token” costs that were not budgeted for (Medium, 2026; Unosquare, 2026).

How to Gain Traction: Responsible AI

To navigate the black ice, businesses must shift from a “move fast and break things” mindset to one of Responsible AI. This isn’t just a buzzword; it is the winter tires and traction control of the digital age.

Responsible AI frameworks prioritise three pillars: Realisation, Reputation, and Regulation (EY, 2025).

1. Data Sovereignty: Ensuring that every AI tool used – official or otherwise – operates under a protected license where data is masked and never used for external model training.
2. Algorithmic Transparency: Moving away from “black box” systems toward models that offer explainability, ensuring that decisions (like hiring or credit scoring) are not biased or discriminatory.
3. Governance by Design: Implementing “AI Gateways” that log all prompts and responses, providing the audit trail necessary to defend business decisions in court or during a regulatory audit.

The dangers of AI for businesses are rarely spectacular explosions; they are quiet, frictionless slips. Shadow AI, unprotected data sharing, and ignored environmental footprints are the patches of black ice that catch even the most sophisticated companies off guard. To survive the transition into an AI-driven economy, businesses must look past the shiny exterior of the technology and invest in the governance structures that keep them on the road.

Need help implementing responsible AI and avoiding the pitfalls that might send your data spiralling out of control? Book a call with our team to learn how we get results – the right way.

References

Espria (2025) Shadow AI: Executive Briefing on Real Risks, Business Impact and Mitigation. Available at: https://www.espria.com/resources/shadow-ai-executive-briefing-on-real-risks-business-impact-and-mitigation/ (Accessed: 30 March 2026).

EY (2025) The business case for responsible AI. Available at: https://www.ey.com/en_us/insights/ai/the-business-case-for-responsible-ai (Accessed: 30 March 2026).

GOV.UK (2025) Report: Water use in AI and Data Centres Executive summary. Available at: https://assets.publishing.service.gov.uk/media/688cb407dc6688ed50878367/Water_use_in_data_centre_and_AI_report.pdf (Accessed: 30 March 2026).

Medium (2026) How the AI Industry Created $644 Billion of Economic Vandalism in 2025. Available at: https://skooloflife.medium.com/how-the-ai-industry-created-644-billion-of-economic-vandalism-in-2025-1ca0d71ab6f2 (Accessed: 30 March 2026).

Online Learning Consortium (2025) The Real Environmental Footprint of Generative AI: What 2025 Data Tell Us. Available at: https://onlinelearningconsortium.org/olc-insights/2025/12/the-real-environmental-footprint-of-generative-ai/ (Accessed: 30 March 2026).

Unosquare (2026) AI Implementation Mistakes That Cost Millions | Avoid These Errors. Available at: https://www.unosquare.com/blog/ai-development-mistakes-that-cost-companies-millions-and-how-to-avoid-them/ (Accessed: 30 March 2026).

TL;DR: As of the Spring ‘26 release, Agentforce Voice has moved from pilot to General Availability (GA), transforming from a simple chatbot with a voice to a fully autonomous, reasoning-capable agent. Powered by the Atlas Reasoning Engine, it can now execute complex tasks (like processing returns or scheduling technicians) without human intervention.

Now GA: Native integration into Service and Sales Cloud, low-code “Canvas” builders for voice logic, and ultra-low latency for natural, real-time conversation.

Coming Soon: An Agent Marketplace for industry-specific voice templates and voice-activated Slackbot agents for internal productivity.

Who Wins: High-volume service centres (scalability), B2B sales teams (automated logging), and regulated industries (secure, AI-governed interactions).

***

The landscape of customer engagement has shifted from passive assistance to autonomous action. Salesforce has formally moved beyond the era of “Copilots” to the era of the Agentic Enterprise. At the heart of this shift is Agentforce Voice, a solution that finally bridges the gap between traditional telephony and autonomous AI.

This article outlines the current state of Agentforce Voice in Spring 2026, what has reached General Availability (GA), and the roadmap for the remainder of the year.

The State of Agentforce Voice: Spring 2026

In the Spring ‘26 release cycle, Salesforce has completed a massive rebranding and architectural overhaul. The most notable change is the evolution of Service Cloud into Agentforce Service and Sales Cloud into Agentforce Sales. Voice is no longer an “add-on” feature; it is a native, intelligent layer integrated into the core of these platforms.

What is General Availability (GA) Now?

As of the February 2026 production rollouts, several key voice capabilities are now GA for all enterprise customers:

• Omni-Channel Autonomous Voice Agents: Unlike previous IVRs (Interactive Voice Response systems) that relied on rigid decision trees, Agentforce Voice agents now use the Atlas Reasoning Engine. Digital Agents can understand natural language, interpret intent and nuance, and execute multi-step tasks – such as qualifying and processing a return or scheduling a field service technician – entirely through voice interaction without human intervention from the organisation.

The result? Consumers must no longer endure waiting in queues, repeating their queries to several agents, and listening to mind-numbing hold music, whilst service teams are free to focus on complex cases – even during seasonal surges!.

• Unified Agentforce Builder (Canvas View): Admins can now build voice-enabled agents using the same low-code builder used for chat and messaging. The new Canvas View allows for the visualisation of branching logic, enabling teams to “drag and drop” voice actions that are grounded in the same Data Cloud and Knowledge libraries as their digital agents. This makes it easier than ever for users to create and manage their digital employees.
• Native Einstein Conversation Insights (ECI) Storage: A major milestone in Spring ‘26 is that ECI data is now stored natively on the Salesforce platform. This eliminates external data silos and allows voice call summaries to trigger Salesforce Flows or Apex actions instantly upon the conclusion of a call.
• Ultra-Low Latency Interaction: This sounds extremely technical, but in simple terms, to prevent the annoying “robotic pause” common in older AI voice systems, Salesforce has optimised the voice stack for near-instant response times, supporting natural interruptions and clarifications – just like a human conversation.

What is to Come: The 2026 Roadmap

Salesforce has signalled a “Summer of Autonomy” with several features currently in pilot or slated for late 2026:

• Agent Marketplaces: Later this year, Salesforce is expected to launch a dedicated section of the AppExchange for pre-built AI agents. This will allow companies to download industry-specific voice agents (e.g., a “Healthcare Billing Agent” or “Retail Returns Specialist”) that come pre-configured with relevant compliance and logic.
• Personalised Slackbot Agents: Currently in pilot for early 2026, Salesforce is moving toward “personal agents” in Slack that can join voice huddles, summarise action items, and update CRM records via voice commands in real-time.
• Advanced Multi-Lingual Nuance: Future updates are focused on expanding “Emotional Intelligence” markers, allowing voice agents to better detect frustration or urgency and initiate an Intelligent Human Handoff with the full transcript and sentiment analysis provided to the live agent.

Who Benefits Most from Agentforce Voice?

The move toward autonomous voice is not just a tech upgrade; it is a strategic shift for specific types of organisations.

1. High-Volume Service Centres (E.g. Retail & Utilities)

Companies dealing with seasonal spikes or high volumes of routine inquiries (e.g., “Where is my order?” or “How do I pay my bill?”) benefit from 24/7 availability. Agentforce Voice allows these companies to scale support without increasing headcount, maintaining a consistent brand voice across every call.

2. Complex B2B Sales Organisations

For sales teams, the integration of voice into Agentforce Sales is transformative. Sales reps no longer need to manually log notes. The system captures the “structured intelligence” of a call, scores the lead against the Ideal Customer Profile (ICP), and automatically creates follow-up tasks. This is particularly valuable for companies with long sales cycles where context is easily lost.

3. Regulated Industries (E.g. Healthcare & Finance)

With the launch of My Trust Centre, Salesforce has provided a tailored view of security and compliance. Organisations in regulated sectors can now deploy voice agents with the confidence that every interaction is governed by the Einstein Trust Layer, ensuring data privacy and biometric security throughout voice authentication.

From Talking to Doing

As of 2026, the “portal-to-ticket” era is effectively over. Agentforce Voice has turned the phone channel – historically a high-cost silo – into a proactive, data-rich asset. By grounding voice interactions in real-time CRM data, businesses are no longer just answering questions; they are completing “jobs to be done” at the speed of conversation.

Is Voice functionality piquing your interest? Book a call with our experts today to find out how you could cut costs, win customers and deliver stellar service with Salesforce.

References

Cloud Analogy (2026) Salesforce Agentforce Trends 2026: The Future of AI-Powered CRM. Available at: https://blog.cloudanalogy.com/salesforce-agentforce-trends-2026-the-future-of-ai-powered-crm/ (Accessed: 1 April 2026).

Codleo (2026) Complete Guide to Salesforce Spring ’26 Release & Agentforce Transformation. Available at: https://www.codleo.com/blog/salesforce-spring-26-release (Accessed: 1 April 2026).

Grazitti Interactive (2026) All About Salesforce Agentforce Voice. Available at: https://www.grazitti.com/blog/turn-customer-conversations-into-trusted-experiences-with-salesforce-agentforce-voice/ (Accessed: 1 April 2026).

Marmato Digital (2025) Agentforce Voice: Deliver Conversational AI Voice Support at Scale. Available at: https://marmatodigital.com/agentforce-voice/ (Accessed: 1 April 2026).

Peergenics (2026) Agentforce Trends for 2026 and Beyond. Available at: https://www.peergenics.com/post/agentforce-trends-for-2026-and-beyond (Accessed: 1 April 2026).

Salesforce (2026) Salesforce Targets the ITSM Status Quo: 180 Organizations Replace Legacy Support Tools with Agentforce IT Service. Available at: https://www.salesforce.com/news/press-releases/2026/02/26/agentforce-it-service-selected-for-itsm/ (Accessed: 1 April 2026).

Salesforce Monday (2026) Agentforce in 2026: What’s New in Salesforce’s Agentic AI Platform. Available at: https://salesforcemonday.com/2026/01/29/agentforce-january-2026-updates-features/ (Accessed: 1 April 2026).

TechForce Services (2026) Salesforce Spring ’26 Release Notes & Agentforce Updates. Available at: https://www.techforceservices.com/blog/salesforce-spring-26-release-guide/ (Accessed: 1 April 2026).

As businesses and marketers, we’ve spent the last few years obsessed with the “how” of Artificial Intelligence – how it can streamline our workflows, how it can cut costs, and how it can help us make more money. But we’ve been so preoccupied with the fear of being ‘left behind’ in this new era that we have spent far too little time asking about the “who.” Who are we actually using this technology for, and do they actually care about our newfound obsession with generative tools? The answer, as I found out from real, everyday consumers this week, is ‘yes’.

We need to stop underestimating the modern consumer. They are becoming more AI-savvy by the day, and their eye for generated content is sharper than we think. A quote in the recent Canva ‘State of Marketing and AI Report, 2026’ really stood out to me and sums it up well:

“AI investment and optimism are accelerating faster than consumer trust.”

This is something I witnessed myself in real time, and it very much got me thinking about how often we consider the consumer amidst all the benefits AI can bring to our organisations. Just yesterday, while scrolling through social media, an UpCircle advert caught my eye. If you aren’t familiar, they are a popular natural beauty brand known for their coffee-based, ‘upcycled’ skincare and ethical ethos.

However, it wasn’t the product that grabbed my attention in the midst of my doom-scroll – it was the comment section. Comments like “Stop using AI slop in your ads” and “I’ve already seen three AI-generated ads from you today” dominated the thread. Consumers weren’t just scrolling on by; they were pausing to express genuine disappointment. Intrigued, I expanded the comment section.

To their credit, UpCircle was present in the comments, responding gracefully. They explained that, as a small team, they are using AI to amplify their brand and compete with the industry giants. And honestly? In my opinion, they should. In 2026, it is becoming nearly impossible to remain competitive without these tools.

But it’s not about if we use it; it’s about how we use them.

Mission VS Machine

What struck me most was the level of awareness from everyday consumers. As a marketer at a Salesforce consultancy, my exposure to the world of AI is likely much broader than that of my fellow consumers, so I almost felt pride in their astute criticism and their demand for better treatment as brand patrons. They weren’t just annoyed by the aesthetics; they were frustrated by the perceived hypocrisy. They questioned how a company built on wellness, ethics, and sustainability could lean so heavily into high-scale AI generation.

When I visited UpCircle’s website, their mission was clear: “Better for you, better for the world.” Now, UpCircle is a BCorp, and have been since 2022, they clearly care about their community and the environmental impact of their activity. Of course they do, that’s how they built their brand in the first place. But we can’t forget that this is indeed a brand, a business. And a business must be efficient; it must survive. This deep dive into my social media on a Monday night really got me thinking about the balance and the use of AI for good. And without spending hours researching, what do I know? UpCircle may completely offset the environmental footprint made by their use of AI. In fact, as a company currently in the process of BCorp certification – I can tell you that this is essentially a requirement.

I’m not judging a small business for using AI – I use it myself, how could I? I work for a company that develops it! But the purpose of writing this isn’t about UpCircle. They are simply doing what every business is doing: trying to survive. And I’ll bet that, as a small BCorp, they are probably doing a lot more than most businesses to negate the environmental impact of their AI use. A quick scroll on their social media pages unearths thoughtful, environmentally conscious, REAL content. They have simply decided to use an LLM to speed up their ad creation.

So, no, this isn’t a criticism of a beauty brand. In fact, they actually appear to be using AI responsibly. I was simply inspired by this comments section to really think about the modern consumer. Yes, there are the people that fall for videos of dogs nursing injured bunnies back to health, or more seriously the slew of AI scams that are popping up at an alarming rate, but the average consumer is much more awake than businesses may think. They are rejecting AI ‘slop’, swiping up on generated content, turning up their noses at unhelpful digital agents.

The damage here isn’t just bad ad performance; it’s the erosion of customer loyalty. When a brand known for “natural” and “human” values pivots to “synthetic” and “automated,” the cognitive dissonance (discomfort caused by contradiction) for the consumer is jarring. If we use AI to free up time in our day-to-day operations, we must have a plan for that time. We shouldn’t use it just to produce more noise; we should use that conserved energy to pour back into the consumer experience.

The brands that will succeed in this era of technology are not the ones that automate the most. They are the ones who use automation to be more human and serve our fellow humans better. We must do right by our audience. Because at the end of the day, if we lose their trust in the pursuit of efficiency, we’ve lost the very thing that makes a brand worth building in the first place.

The brands that keep their customer, and subsequently humans, at the centre are the ones that will win the AI race, and not the other way around.

Lauren Tovey – Marketing Manager – Performa

TL;DR UK energy and water companies are currently trapped between £4.48 billion in consumer debt, a 10-year grid connection backlog, and severe skills shortages. Traditional manual processes are no longer fast enough to manage these compounding crises.

For Salesforce users, the solution lies in Agentforce and Slack OS, which provide a “digital workforce” to automate complex actions:

• Debt Recovery: Agentforce uses autonomous reasoning to identify at-risk customers early and set up personalised, compliant repayment plans.
• Grid Efficiency: Slack OS serves as a “command centre,” using AI agents to coordinate between engineers and planners, removing the administrative friction that delays infrastructure projects.
• Bridging the Skills Gap: AI “co-pilots” support field technicians with instant access to technical data, effectively increasing operational capacity without needing new hires.

By shifting from simple automation to agentic technology, utility providers can reduce bad debt, accelerate the green transition, and restore public trust through better transparency.

Unsurprisingly, the UK energy and utilities sector is currently navigating unprecedented operational and financial strain. Recent data from Ofgem reveals that domestic consumer energy debt reached a record high of £4.48 billion by mid-2025, a staggering 71 per cent increase since 2023 (Ofgem, 2025). This financial burden is compounded by a 10-year waiting list for grid connections and a critical shortage of skilled engineers. For many providers, traditional manual processes and siloed data are no longer sufficient to manage these escalating crises.

As the industry reaches a breaking point, a new generation of agentic technology is emerging as the primary solution. Salesforce’s Agentforce and the Slack Operating System (OS) are shifting the paradigm from basic automation to autonomous action, providing the “digital labour” required to stabilise the market and restore public trust.

Solving the Debt Crisis with Compassionate Intelligence

With nearly three-quarters of the £4.48 billion debt stack held by customers with no repayment plan in place, the pressure on collections teams is immense (Ofgem, 2025). Traditional “one-size-fits-all” debt collection often exacerbates customer vulnerability and leads to costly regulatory interventions.

Agentforce offers a sophisticated alternative. These AI agents do not merely send automated reminders; they possess the “reasoning” capabilities to analyse a customer’s specific billing history, energy usage patterns, and external socio-economic factors. When a customer interacts with a provider, an Agentforce agent can proactively identify those at risk of falling into arrears before the first payment is missed.

By integrating with the Einstein Trust Layer, these agents can safely suggest personalised repayment plans or recommend government-backed support schemes, such as the Debt Relief Scheme. This ensures that the 20 per cent to 40 per cent of debt attributed to “move-in” anomalies (where customers use energy without setting up an account) is captured and managed through autonomous, plain-language conversations (Ofgem, 2025).

Slack OS: Breaking the Gridlock

The infrastructure bottleneck is a problem of coordination as much as capacity. Building a greener grid requires seamless collaboration between developers, planners, and site engineers.

Slack OS acts as the “front door” for these complex projects. By treating Slack as a work operating system rather than a messaging app, utility companies can create “Project Huddles” where Agentforce agents sit alongside human teams. When a new renewable project applies for connection, an agent can automatically pull data from the Salesforce Energy and Utilities Cloud, summarise technical requirements, and alert the relevant planning team in a dedicated Slack channel.

This eliminates the “context switching” that often delays critical infrastructure decisions. With Salesforce records now deeply integrated into Slack, over 6.4 million records are shared weekly across the platform, allowing engineering teams to approve site designs or update project statuses without ever leaving the conversation (Salesforce UK, 2026).

Bridging the Skills Gap in the Field

The UK water and energy sectors are facing an acute shortage of authorised engineers, a problem that threatens the delivery of the £104 billion investment programme required for the water industry (Bionic, 2026). When experienced staff are stretched thin, the risk of operational errors, such as sewage spills or grid failures, increases significantly.

Agentforce 360 for Field Service transforms how limited human resources are deployed. Instead of dispatchers manually juggling schedules, AI agents use the Atlas reasoning engine to assign technicians based on real-time location, specific skill sets, and job priority.

For on-site technicians, the technology acts as a digital co-pilot. A technician facing a complex repair on an ancient water main can use Agentforce in Slack to query technical manuals, view historical asset data, and even draft close-out notes using voice commands. By automating these administrative burdens, utility companies can effectively increase their “boots on the ground” capacity by 20 per cent or more without increasing headcount, ensuring that the most critical repairs are handled by the right person at the right time.

Restoring Trust through Transparency

In an era of intense public scrutiny regarding water pollution and energy pricing, transparency is the ultimate currency. The 2026 Water Reform Bill has made it clear that “self-monitoring” is no longer an option for utility providers.

Agentic technology provides an immutable audit trail. Every decision made by an agent is grounded in the company’s internal data and knowledge base, ensuring compliance with evolving regulations. In the event of a service failure or environmental incident, Slack OS provides a centralised hub for “case swarming,” where experts can collaborate to resolve the issue while the AI agent provides real-time updates to affected customers and regulators. This proactive communication is essential for rebuilding a reputation that has been damaged by years of perceived opacity.

Conclusion: The Future of Digital Labour

The challenges facing UK energy and utilities are no longer manageable through incremental efficiency gains. The £4.48 billion debt mountain and the gridlock crisis require a fundamental shift in how work is done. At Performa, we have witnessed first-hand how, by deploying Agentic technology like Agentforce and Slack OS, companies are not just buying software; they are onboarding a digital workforce capable of solving complex problems at scale.

This technology allows human employees to move away from the “drudgery” of data entry and manual scheduling, focusing instead on the high-value, empathetic work that a machine cannot do. For the UK utility sector, the choice is clear: embrace agentic technology or risk being left behind in a gridlocked and indebted past.

Book a call with us today to discover how Agentforce can transform your operations.

References

Bionic (2026). What recent energy debt trends tell us about the health of UK businesses. [online] Available at: https://bionic.co.uk/business-energy/guides/billing-payments/what-recent-energy-debt-trends-tell-us-about-the-health-of-uk-businesses/ [Accessed 20 Mar. 2026].

Ofgem (2025). Debt strategy update: supporting the reduction of energy debt. [online] Available at: https://www.ofgem.gov.uk/policy/debt-strategy-update-supporting-reduction-energy-debt [Accessed 20 Mar. 2026].

Salesforce UK (2026). Agentforce for Energy & Utilities. [online] Available at: https://www.salesforce.com/uk/energy-utilities/artificial-intelligence/ [Accessed 20 Mar. 2026].

Salesforce UK (2026). Slack is the work operating system for the agentic enterprise. [online] Available at: https://www.salesforce.com/uk/slack/ [Accessed 20 Mar. 2026].

TL;DR 82% of constituents feel ignored by government digital services, largely because many departments are still running on 50-year-old legacy systems. This technical debt wastes roughly 30.6 million hours a week in the UK alone.

The Solution: Moving away from “siloed” tech to a Single Source of Truth. By using Slack as an Agentic Operating System, PubSec teams can use a conversational, mobile-friendly interface to manage everything from HR to IT without needing to raise constant support tickets.

The “Agent” Advantage: We leverage Agentforce technology to:

• Triage Case Work: Like identifying and scheduling pothole repairs via photo analysis and GPS.
• Manage Grants: Slashing processing times from months to weeks by automatically pulling data from multiple charities and trusts.
• Streamline Hiring: Using AI to surface the best candidates from thousands of applications, saving hundreds of hours for HR teams.

The Future: With only a small fraction of government entities currently on Salesforce, we expect a massive migration toward these AI-driven, “Agentic” models to meet modern constituent expectations.

The divide between public expectation and digital reality has reached a breaking point. According to the Salesforce Connected Government Report, a staggering 82% of constituents feel that their government does not prioritise the customer experience (Salesforce, 2024). While the private sector has spent the last decade perfecting frictionless, “one-click” interactions, the public sector (PubSec) is often found wading through a quagmire of legacy infrastructure.

In some departments, the core technology in use is often, let’s be honest, not shockingly over 50 years old. These “green-screen” systems and siloed databases do more than just slow down internal operations; they create a fundamental disconnect between the state and the citizen. When technology is this archaic, data stays trapped in departmental vacuums, leading to repetitive paperwork, long waiting times, and a general sense of frustration. To fix the experience for the constituent, we must first fix the operating system of the government.

What is Missing from PubSec?

The missing ingredient in modern public service is not just “new tech,” but connectivity. While connectivity is often dismissed as a corporate buzzword, in a PubSec context, it refers to the tangible benefits of an integrated “single source of truth.” When systems do not talk to one another, the cost is felt by everyone: the exhausted front-line team, the frustrated user, and the wider organisation struggling with budget cuts.

The scale of this inefficiency is quantifiable. It is estimated that 30.6 million hours a week are wasted across the UK public sector workforce due to systemic inefficiencies (Civil Service World, 2024). These hours are lost to manual data entry, “swivel-chairing” between incompatible software, and creating workarounds for legacy workflows that were never designed for the digital age.

Historically, government departments have attempted to bridge these gaps with rigid, monolithic Enterprise Resource Planning (ERP) systems. However, these often become “digital concrete” – hard to change, expensive to maintain, and requiring a formal IT support ticket for even the smallest adjustment. What PubSec truly needs is a secure, flexible operating system that empowers staff rather than hindering them.

The New Slack OS: The Command Centre for PubSec

The evolution of Slack from a messaging app to a comprehensive “Operating System” (OS) enhanced by Slack AI is a game-changer for the public sector. As Steve Hamrick, VP of Product Management at Salesforce, describes it, the goal is to make work “fast, easy and delightful” (Salesforce, 2025).

For a government worker, the Slack OS provides:

• Universal Accessibility: The ability to move seamlessly between mobile and desktop ensures that field workers, such as social workers or building inspectors, stay connected to the hub.
• Low-Code/No-Code Agility: Teams can build their own workflows. You no longer need to wait six months for a developer to automate a simple approval process.
• A Conversational Interface: Rather than navigating complex menus, users can simply ask Slack AI to find a policy document or summarise a long thread of case notes.

Slack is now being positioned as the “command centre” for Salesforce. By integrating these platforms, PubSec organisations can finally move away from fragmented tools and into a unified environment where all apps are connected in one conversational interface.

Expanding the Salesforce Ecosystem in Public Service

Salesforce has spent years refining a suite of products specifically tailored for the unique pressures of the public sector. This timeline of innovation has moved beyond simple CRM to include:

• Experience Cloud: Providing external engagement portals for citizens to track their own applications.
• Grant and Benefits Management: Streamlining the disbursement of vital funds.
• Public Health and Case Management: Ensuring vulnerable individuals do not fall through the cracks of a disjointed system.
• Facilities Ops and Licensing: Managing the physical and legal infrastructure of our cities with specialised AI for every task.

The Rise of the Agentic Organisation

The next frontier is the Agentic Organisation. This moves beyond simple automation to use “Agentforce” – autonomous digital agents capable of reasoning and executing complex tasks. At Performa, for example, we have developed specific accelerators for digital agents that are already transforming key service areas.For PubSec, these look like:

1. Case Triage: Road Maintenance

The state of UK roads is a frequent point of national contention; one might say our tarmac is crying out for a digital intervention. Imagine a constituent reporting a pothole. Instead of a static form, a Highway Agent engages them via Slack or a mobile portal:

• The Agent asks for a photograph and uses GPS coordinates to pinpoint the location.
• The reasoning engine analyses the image, comparing the pothole to the road’s centre lines to estimate size and depth.
• It then triages the report, automatically scheduling a repair crew based on priority and existing assignments, all while keeping the constituent updated in real-time.

2. Grant Management & Disbursements

Consider a healthcare trust identifying a patient who needs energy-saving home improvements. A Grants Agent can collect data from the trust, the non-profit, and third-party energy providers simultaneously. By using its reasoning engine to determine eligibility and priority, the Agent can allocate funds in weeks rather than months, removing the “paperwork wall” that often stops help from reaching those in need.

3. Candidate Qualification

Government entities often struggle with high staff turnover and a deluge of applications. An Agent can be deployed to surface only the most qualified candidates. In advanced stages, agents can even analyse unstructured data from applicant videos to identify the top five candidates based on job descriptions and call scripts. This saves hundreds of hours for underfunded HR teams, eliminating the need for expensive external consultancies.

A Future of Efficiency

All these processes can be overseen from the Agentic OS in Slack. There is no need to switch between dozens of tools; everything is handled within a conversational interface where knowledge is centralised.

We are on the cusp of a major shift. In the US, there are currently between 1,000 and 1,500 public-sector Salesforce (Salesforce, 2025) customers among nearly 100,000 total government entities (Census of Governments, 2026). However, as the “Agentic” model proves its worth, we predict these numbers will climb rapidly. The UK is set to mirror this trend as local and central government bodies seek to reclaim those 30.6 million wasted hours.

The transition from legacy “dead-end” tech to an Agentic Operating System is no longer a luxury – it is a necessity for a functioning modern democracy.

Are you ready to reclaim your team’s time and transform the constituent experience?

Book a call with our experts today to explore our PubSec accelerators.

References

Civil Service World (2024). The Productivity Gap: Measuring Inefficiency in the UK Public Sector. [online] Available at: https://www.civilserviceworld.com [Accessed 18 Mar. 2026].

Salesforce (2024). Connected Government Report: Global Trends in Public Sector Digitisation. 4th ed. San Francisco: Salesforce Research.

Salesforce (2025). The Future of the Slack OS: Insights from Steve Hamrick. [online] Salesforce Newsroom. Available at: https://www.salesforce.com/news [Accessed 19 Mar. 2026].

Salesforce (2026). Webinar: Reimagine Mission Delivery with an Agentic Operating System for Public Service. [online] Available at: https://www.salesforce.com/events/webinars/reimagine-mission-delivery-with-an-agentic-operating-system/.

U.S. Census Bureau (2026). 2026 Census of Governments: Preliminary Report on Digital Infrastructure. Washington, D.C.: Government Printing Office.

In the world of enterprise technology, the term “legacy” is often used as a polite euphemism for “the thing that keeps the CTO up at night.”

As a Salesforce consultancy, we see this daily. Organisations are tethered to ageing, monolithic systems that were cutting-edge when installed but have since become anchors, slowing innovation and draining IT budgets.

Phasing out legacy technology is not just a technical upgrade; it is a strategic necessity. However, the “rip and replace” approach is often too risky. The most successful transitions follow a disciplined, phased programme that balances business continuity with modern agility.

Step 1: The Audit

Before a single line of code is moved, we must understand the current state of our ecosystem. Many legacy systems survive because their full range of dependencies is unknown. Before proceeding with the migration, we must:

• Inventory Every Dependency: Identify and document which apps, databases, and third-party tools pull data from the legacy system.
• Identify Business Logic: Document the unique rules and workflows embedded in the old system. Often, these rules result from years of specific business requirements that must be translated into the new environment.

Step 2: Strategic Triage

Not every piece of legacy tech needs to be treated the same way. We recommend categorising your assets using the “Retire, Retain, or Replace” framework.

Step 3: The Data Cleansing Ritual

Data is the most valuable asset in your legacy system, but it is also likely the messiest. Moving “dirty” data into a clean, modern platform like Salesforce is like putting an old, leaking engine into a brand-new car. Data cleansing consists of three core steps:

1. Classification: Categorise data based on its functional importance and regulatory requirements (GDPR/SOX).
2. Deduplication: Remove the thousands of duplicate records that have inevitably accumulated.
3. Archiving: Do you really need twenty years of transaction history in your active CRM? Move historical records to a low-cost, secure data lake where they remain accessible for audits but don’t clutter your new environment.

Step 4: To Scream or Not to Scream

The “Scream Test”, the correct approach or a horror movie waiting to happen? 

The “scream test” is a blunt-force method of infrastructure management where a server or service is intentionally disabled to see if any users or dependent systems “scream” in protest (an apt taxonomy). While effective at identifying forgotten dependencies in a chaotic environment, it is inherently disruptive and can carry significant risk to business continuity. 

Modern, data-driven organisations are increasingly moving away from this “trial by fire” approach in favour of observability and automated discovery.

By leveraging Automated Monitoring and Resource Usage Analysis through tools like Azure Advisor or Datadog, IT teams can identify “zombie” resources based on actual traffic patterns rather than guesswork. 

For those who still require a validation step, a “Soft” Shutdown – such as disabling network connectivity or stopping a specific application pool – provides a safety net, allowing for near-instant restoration if a critical dependency is revealed. 

Furthermore, Data-Driven Discovery through service-mapping tools such as Faddom or ServiceNow ensures that dependencies are mathematically mapped before a single switch is flipped. 

Ultimately, shifting toward a slightly less excitingly named Policy-Based Management approach, where “Cliff Edge,” preemptive announcements and mandatory owner tagging are the norm, replaces the panic of a scream test with a predictable, transparent decommissioning process. See below for our soft shutdown checklist. 

The Brownout Checklist 

Pre-Shutdown: The “Safety Net”

• [ ] Baseline Performance Logs: Record current traffic and connection counts. This provides a “normal” benchmark to compare against after the shutdown.
• [ ] Identify Stakeholders: Notify department heads of the scheduled brownout window. Clearly define the “Scream Window” (e.g., 4 hours) where IT will be on high alert.
• [ ] Verify Rollback Credentials: Ensure that at least two administrators have local access and the necessary permissions to instantly re-enable services or network ports.
• [ ] Snapshot/Backup: Perform a final virtual machine snapshot or full data backup. Even though you aren’t deleting data yet, a soft shutdown can sometimes trigger unexpected “fail-close” states in connected apps.

Execution: The “Soft” Disruption

• [ ] Isolate the Network: Instead of a hard power-off, disable the specific Virtual LAN (VLAN) or firewall rule. This simulates a server failure for users but keeps the server’s internal state active for easy diagnosis.
• [ ] Stop Application Pools: If it’s a web server, stop the IIS or Apache service first. This allows the OS to remain reachable for pings/management while the “service” appears down to users.
• [ ] Monitor Connection Refusals: Watch your load balancer or firewall logs. An immediate spike in “Connection Refused” errors from a specific internal IP address identifies a hidden service account or API dependency.

Post-Shutdown: Validation and Triage

• [ ] Monitor Helpdesk Tickets: Designate a specific tag in your ticketing system (e.g., #LegacyMigration) to catch reports related to the shutdown immediately.
• [ ] The “Wait and See” Period: Maintain the soft shutdown for a full business cycle (usually 24 to 48 hours) to account for daily or overnight batch jobs that only run once a day.
• [ ] Document “The Scream”: If someone complains, don’t just turn it back on. Use the disruption to identify the exact service, user, or process that was missed in the audit phase and document it in your CMDB.

Final Decommissioning

• [ ] Final Sunset: If no “screams” occur after a full week of network isolation, proceed to a hard power-down.
• [ ] Physical/Logical Purge: After 30 days of “Hard Off” with no issues, you can safely reclaim the hardware resources or delete the VM.

Phase 5: The Incremental Migration (The “Strangler” Pattern)

The biggest mistake companies make is the “Big Bang” rollout. Instead, we advocate for the Strangler Fig Pattern. Much like the vine that grows around a tree and eventually replaces it, you should gradually migrate functional modules one by one.

• Start with a “Quick Win”: Choose a single high-impact but low-complexity process, such as lead management or customer service ticketing.
• Build an Integration Layer: Use APIs or middleware to allow the new system and the legacy system to “talk” to each other during the transition.
• Sync in Real-Time: Ensure that as users move to the new system, their data is reflected back in the legacy database (and vice-versa) until the old system is ready to be switched off.

Phase 6: Change Management and Adoption

Transforming technology is (or should be) straightforward when the right processes are in place; transforming people and ways of working is the hard part. Legacy systems often have “super-users” who know every obscure shortcut. Replacing their familiar tool can lead to resistance. For successful, sustainable transformation, leaders must incorporate the following on their roadmap:

• Hands-on Training: Do not just provide manuals. Host workshops where users can see how the new system speeds up their repetitive daily tasks and/or enhances their work.
• User Feedback Loops: Involve end-users in the testing phase. If you are involved in building the new system, you are far more likely to champion its adoption.

Phase 7: Logical and Physical Decommissioning

Once the new system is fully operational and the data has been verified, it is time to pull the plug.

• Final Validation: Perform a final backup and ensure that every business object has been accounted for.
• Deactivate Access: Disable user logins first, followed by the server infrastructure.
• Contract Wind-down: Coordinate with procurement to end maintenance and licensing agreements, finally realising the cost savings promised at the start of the project.

In Summary

Phasing out legacy technology is a marathon, not a sprint. By moving away from monolithic, high-maintenance systems toward flexible, API-driven platforms like Salesforce, organisations can finally stop spending 80% of their budget on maintenance and start investing in growth.

The goal isn’t just to have a “new” system; it’s to have a system that is capable of evolving as fast as the market does.

Ready to Modernise?

If your legacy systems are holding your business back, you don’t have to navigate the transition alone. We specialise in helping organisations move from technical debt to digital agility. Need guidance on your transformation? Get in touch with our experts today.

References

• AlterSquare (2025) Legacy System Modernization: A Step-by-Step Guide. Available at: https://altersquare.medium.com/legacy-system-modernization-a-step-by-step-guide-1b8b0acde6c6 (Accessed: 22 January 2026).
• Future Processing (2025) Legacy system modernisation: challenges and common approaches. Available at: https://www.future-processing.com/blog/legacy-system-modernisation/ (Accessed: 22 January 2026).
• Hyland (2024) How & When to Modernize Legacy Systems. Available at: https://www.hyland.com/en/resources/articles/legacy-system-modernization (Accessed: 22 January 2026).
• Neev (2025) Legacy System Decommissioning: Retiring Systems Without Losing Critical Data. Available at: https://neevdata.com/blog/legacy-system-decommissioning-retiring-systems/ (Accessed: 22 January 2026).
• Opal Solutions (2026) Legacy Modernisation: Rewrite, Refactor or Retire Systems. Available at: https://opaltechsolutions.co.uk/post/legacy-modernisation-when-to-rewrite-refactor-or-retire-legacy-systems (Accessed: 22 January 2026).
• Stefanini (2024) How to Modernize Legacy Applications: A Step-by-Step Guide. Available at: https://stefanini.com/en/insights/news/how-to-modernize-legacy-applications-a-step-by-step-guide (Accessed: 22 January 2026).