The missing piece in every UK AI productivity conversation

The UK is pushing hard on AI. London’s new AI and Jobs Taskforce shows how quickly the debate has moved from technology adoption into jobs, productivity, and the future of work. That focus is important, but senior leaders should be careful not to mistake AI ambition for organisational readiness.

 

The real productivity question is whether a business can carry context across teams, systems, and customer interactions. When the rationale behind a decision disappears into a channel, or a customer commitment sits apart from the work needed to fulfil it, AI has little stable ground to build on. It can accelerate activity, but it cannot repair the underlying operating memory.

 

 

AI adoption is not the same as productivity

The pressure to adopt AI is growing. UK government AI Adoption Research found that around one in six businesses are currently using at least one AI technology.

 

The business case remains uneven. PwC’s 29th Global CEO Survey found that 30% of CEOs say their company has realised additional revenues from AI adoption over the past year, while 26% say costs have decreased. However, 56% have seen neither higher revenues nor lower costs.

 

The issue is not simply whether companies are using AI, but whether the operating model can turn it into performance. When decisions are scattered across channels - account records, support tickets, and people’s memories - AI can add speed without adding clarity. The organisation may process more activity while still struggling to explain the rationale, ownership, next step, and customer commitment.

 

 

The hidden cost of fragmented work

Most collaboration systems were built to help people communicate. They were not designed to preserve institutional memory. A customer issue rarely belongs to one team. It may begin in support, touch sales, involve product, require engineering input, and carry commercial implications. A decision may be discussed in a meeting, referenced in a channel, logged in a ticket, and later explained to a customer. The work is connected, but its evidence is often scattered.

 

When those relationships are left implicit, people become the integration layer. Employees search for answers that already exist, managers reopen discussions because the rationale has disappeared, and customers repeat themselves when context is lost between handoffs.

 

The problem is not only inefficiency; it also weakens accountability. This matters particularly in the UK context, where the ONS has consistently flagged weak productivity growth as one of the country’s most persistent economic challenges. Senior leaders often look for productivity gains through process automation, but many delays still occur during context reconstruction. People are not only doing the work; they are constantly rebuilding the conditions needed to do it.

 

 

AI can accelerate the wrong operating model

AI disappoints when it is introduced into a fragmented organisation and simply moves fragmented work faster. An AI system can summarise a document, retrieve a message, suggest a response, or trigger a workflow. Those capabilities are useful. But if the system cannot understand how a customer issue relates to the account history, product decision, engineering discussion, commercial commitment, and eventual outcome, it is still acting on fragments.

 

For business leaders, the risk is not only poor AI output, but misplaced confidence in systems that appear more coherent than they are. AI can support decision-making, but only if it is grounded in the right organisational context and accountable to the people who understand the consequences.

 

 

Shared memory is becoming business infrastructure

The next stage of enterprise AI should be treated as an operating model question. A support ticket should carry its relationship to the customer account, the product issue, the engineering history, the decision rationale, and the service commitment. A new team member should not need months of messages to understand why a choice was made. A leader should be able to see how a decision connects to risk, revenue, customer experience, and execution.

 

That requires shared memory - a structured, permission-aware foundation that moves conversations and decisions into the work objects they affect, rather than leaving them scattered across channels.

 

This is where knowledge graph-based collaboration becomes commercially relevant. The value is not the graph itself, but the way it makes relationships explicit so people and AI systems can work from the same operating context.

 

For AI to deliver business value at scale, that operating context must support three outcomes: precision, efficiency, and safety. Precision comes from grounding answers in authoritative business data and preserving the relationships between decisions, customers, work, and outcomes. Efficiency comes from reasoning against an existing organisational memory rather than repeatedly reconstructing context through search, retrieval, and human intervention. This results in lower token usage and, in turn, lower costs. Safety comes from ensuring AI operates within the same permissions, controls, and accountability structures that govern the rest of the business.

 

 

Governance must be designed into the work

As AI moves closer to action, governance becomes a design problem. In my view, it is where most organisations are currently most exposed. It cannot sit outside the organisation as a policy document reviewed after deployment.

 

Governance is ultimately how organisations create safety at scale. It determines not only what AI systems can access, but also what they can influence, what they can automate, and how their actions can be monitored and reversed when necessary.

 

The CIPD has found that 63% of people would trust AI to inform important decisions at work, but only 1% would trust it to make those decisions. That gap is a design requirement. If AI is helping route work, inform decisions, surface risks, or support customer responses, leaders need to know what it can see, what it can do, what it should hand back to people, and how its actions are recorded.

 

Permission-aware memory matters because productivity without accountability is not a business gain. Precision without safety creates risk. Safety without efficiency limits value. Speed without structure accelerates risk as much as it accelerates work. Companies that get value from AI will fix the memory layer beneath it, preserving context where work happens, retaining decision rationale, and making customer commitments visible across teams.

 

AI may help businesses move faster. The leadership challenge is ensuring the organisation remembers enough for that speed to translate into better judgement, not just more activity. For the UK to close its productivity gap through AI, that leadership challenge cannot be optional. 

 

---

 

Ahmed Bashir is CTO at DevRev

 

Main image courtesy of iStockPhoto.com and Shinsei Motions