# As leaders leave, councils must protect institutional memory > Source: > Published: 2026-08-12 13:25:35+00:00 ##### As reorganisation sees many councils merge into new unitary authorities, the people who could catch AI’s mistakes could be pushed out as part of the process, according to Katryna Peart More than 100 councils across England are collapsing into new unitary authorities by 2028. Under the Staffing Regulations that govern the process, chief executives of predecessor councils aren’t covered by TUPE (Transfer of Undertakings, Protection of Employment) — the rule that normally guarantees staff keep their jobs, on the same terms, when their employer changes. If a chief executive doesn’t secure the top job in the new authority, they’re among the most exposed to being cut — and chief executives typically hold a large share of an organisation’s institutional memory and day-to-day operational judgment. Councils aren’t just losing people. They’re rebuilding the institutional record at the same time. Merging authorities means merging Revenues and Benefits, Housing, Adult Social Care and SEND (special educational needs and disabilities) systems, plus waste management, that were never designed to talk to each other — and officials who’ve done this before say the two hardest parts are the same: knowing what the data actually means, and having the right people left to explain it. The Ministry of Housing, Communities and Local Government (MHCLG) — the department overseeing the reorganisation — launched its own Local AI programme in March 2026, meant to help councils adopt AI responsibly during exactly this transition. Any AI tool deployed against local authority systems will draw on [the rebuilt record](https://www.publictechnology.net/2026/06/09/communities-housing-and-planning/what-uk-councils-stand-to-lose-when-ai-enters-a-reorganisation/) to make sense of what’s left once systems are combined — inheriting whatever gaps, distortions, or omissions that record carries. The record becomes the primary source of truth at the exact moment the people who could correct it are leaving. AI doesn’t need a complete record to produce something that reads like one. Even a stripped-down or heavily compressed document leaves remnants: a name, a date, a fragment to work from. Institutional knowledge that lives in a person doesn’t work that way. Once that person walks out the door, it’s gone entirely — and it’s usually the stuff nobody even notices is greasing the wheels: the workaround nobody wrote down, the reason a decision went one way and not another. Normally, that’s survivable. A colleague, a caseworker, a fellow officer who was actually in the room can say: ‘no, that’s not what happened’, or ‘that’s not the whole story’. The correction depends on someone with real memory of the decision still being around to make it. The same reorganisation that is removing these people is removing that possibility, on the same deadline. The record becomes the primary source of truth at the exact moment the people who could correct it are leaving. This isn’t a general public-sector exodus. Civil service turnover fell to 7.1% in 2024/25, down from 7.5 the year before; resignations dropped from 4.2% to 3.6%. Fewer people are leaving the public sector than a year ago. Against that falling baseline, the churn concentrated in reorganising councils isn’t background noise. It’s a spike, with a specific cause, arriving on a specific date, for a specific set of people — the ones most likely to know when the record is wrong. **Suffolk and Chesterfield** In February 2026, Suffolk County Council confirmed its chief executive Nicola Beach would leave her post at the end of March, after eight years in the job. Rather than recruiting a replacement, councillors folded her responsibilities into two existing executive director roles — a move projected to save roughly £92,000 a year, in the same two-year window the council’s leader called the most critical in its history, with reorganisation, devolution and financial sustainability all landing at once. The authority’s elected leader, councillor Matthew Hicks, said that the departing CEO’s “expertise and love of Suffolk will be missed.” Institutional knowledge, named as a loss, in public, while the reorganisation Beach would have carried context on was still ongoing. At Chesterfield Borough Council, meanwhile, chief executive Huw Bowen retired in May 2026. Announced in December 2025, his successor isn’t an outside hire. It’s his deputy of six years, Christine Durrant, already on the senior team since 2019, chosen explicitly, in the council leader’s words, to “carry forward the respected legacy” of the outgoing chief executive at a moment of major change. The approach is structurally different from Suffolk’s: a direct internal successor, chosen deliberately, with enough runway for a handover to actually happen. That’s not a guarantee the knowledge transfers cleanly — but it’s a plan. Most councils in the middle of reorganisation won’t manage it like Chesterfield. One council chief executive put it bluntly: “We’re all just head down in the MHCLG directive, no one is lifting their heads, there’s no excitement.” That’s an organisation with its full attention on the structural change itself — boundaries, staffing formulas, legal paperwork — and none left over for the fact that its most experienced people are leaving in the middle of it. **Document it while there’s still time** Chesterfield treated succession as a task with its own named owner and its own timeline, separate from the mechanics of the merger. Suffolk didn’t recruit a direct replacement at all — the role was folded into two existing jobs. Every major procurement has an owner, a budget, milestones and formal approval. Institutional memory transfer usually has none of those things. Before reorganisation reaches a senior departure, someone should already be documenting what that person knows that exists nowhere else. Who’s receiving it? By when? None of this is sentimental. Institutional memory is what lets a new authority explain its own decisions instead of guessing at them — the difference between an organisation that knows why it works the way it does, and one that’s just inherited a filing cabinet. *Katryna Peart (pictured above right) is an AI historiographer and institutional memory researcher. She is the developer of Civic Pair Prompting (CPP), a replicable framework for evaluating municipal AI deployments, and the originator of institutional corpus stress-testing methodology for civic documents. She has conducted evaluation work with Google and Uber and her research has appeared in Governing Magazine, Route Fifty, and PM Magazine. Find out more at govai.expert.com.*