What 90 days of AI implementation actually delivers.
A straight answer for UK founders sizing up an AI rollout. Ninety days, well run, gets you measurable outcomes in a £5-50m business. Not transformation. Something better: numbers you can put in front of the board.
Every founder asks the same two questions on the first call. How long is this going to take, and what do I actually see at the end of it. The honest answer is 90 days for a real result in a UK mid-market business, and the result is measurable and small at first. Nobody serious is promising transformation.
What 90 days of AI implementation should produce
By the end of a well-run 90-day engagement, a UK business between £5m and £50m of revenue should have two or three workflows running in production, hundreds of hours of senior time back each month, one or two hires deferred, and decisions arriving in minutes that used to take days. Four numbers. Everything else is decoration.
Live means something specific. Not "a tool somebody is trialling". Running in production against real volume, with the audit trail your finance team needs, a named owner, and a review cadence that catches the edges. The two or three workflows that make it through are the ones already costing the most hours every month: reconciliations, ticket triage, reporting pulls, supplier onboarding, returns matching. The obvious candidates, properly scoped and properly built.
The hours come back from there. A reconciliation that ate 60 hours from a senior finance manager becomes five. A triage queue that needed two support analysts needs one, with the second redeployed onto retention work. Aggregated across two or three live workflows, the range lands consistently in the same place.
Hours back is the number that funds everything else. Once they are back, the "next hire" in ops or finance, mid-level and roughly £42,000 fully loaded outside London, gets shelved. One deferred hire pays back a year-one engagement in roughly fourteen months. Two pays it back in seven. That arithmetic is why most founders, once they have run one of these, run another.
Why most attempts don't get there
The pattern is consistent. Founders who DIY end up with shadow IT and stalled tools in browser tabs. Founders who hire the wrong help end up with a deck of recommendations and no production builds. Bain's Technology Report 2026 finds 90% of leaders say their data foundations are not fit to scale AI. McKinsey's State of AI Trust 2026 reports 88% of organisations using AI but only 23% having scaled a single workflow. The pilots get built on data nobody has stress-tested, and they fail the moment volume goes up.
The work gets sized before the question is answered. The businesses that produce the outcomes on the board above are the ones with someone in the room who has run this thirty times before, who counted the hours properly, picked the workflows that actually move the number, and built them in production rather than in a deck. That capability has to come from somewhere.
The realistic shape of what happens
First workflow live.
One agent in production with audit logging. Hours get counted properly, often for the first time. Baseline measurements captured.
Two more, plus tuning.
Reporting layer added. The first workflow gets its edge cases fixed. The conversation in finance shifts from "is the report ready" to "what's it telling us".
Result in front of the board.
Hours saved totted up. Hire deferral confirmed. Decision lag in two functions measurably shorter. The board meeting moves on to which workflow is next.
Ninety days, well run, gets you measurable outcomes. Not transformation. Numbers you can put in front of the board.
What the spend looks like
Mid five-figure year-one spend is the credible range for a UK business between £5m and £50m of revenue. That covers a Diagnostic to scope the workflows, two or three production builds, and bundled support to keep them running and tuned. Anything materially below is selling you a chatbot. Anything materially above is selling you scope you don't need yet. McKinsey's State of AI 2026 puts median payback on properly scoped production AI workflows inside twelve months. Workflow-by-workflow payback runs faster or slower depending on shape.
Compare the implementation number to one mid-level UK ops hire at £42,000 fully loaded outside London, plus three to six months of onboarding lag before they are productive, plus the churn risk that comes with the role. The arithmetic is not subtle, which is why the conversation is shifting.
Why this matters in 2026
DSIT's January 2026 research puts only 15% of UK SMEs as having deployed AI. BCC and Atos in March 2026 reported 54% of UK firms using AI, but mostly as a support layer rather than inside core workflows. BCG's AI Radar 2026 reports 90% of CEOs expecting agentic ROI this year. The UK businesses that cross to the operating-with-AI side of that gap during 2026 will set the cost-base norms for their sector. The ones that don't will be explaining a thirty-person back office in 2028, against competitors running the same revenue line with eighteen.
Frequently asked questions
How long does an AI implementation take for a UK SME?
Ninety days is the realistic horizon for a first measurable result in a UK business between £5m and £50m of revenue. That means two or three workflows running in production, with audit trails and a named owner. Material P&L impact takes longer, typically six to twelve months once the first builds are live and the team has adjusted around them. Anyone quoting faster than 90 days is either selling a chatbot or hasn't scoped the work.
What does a 90-day AI implementation cost in the UK?
For a UK business in the £5-50m revenue range, mid five-figure year-one spend is the credible range. That covers a Diagnostic to scope the workflows, two or three production builds, and bundled support to keep them running. Compare that to one mid-level UK operations hire fully loaded at roughly £42,000 a year before onboarding lag and churn risk. The arithmetic favours the implementation in almost every case Clerq has scoped.
What workflows should we automate first?
The ones already costing the most hours every month. Reconciliations, ticket triage, reporting pulls, supplier onboarding, returns matching. Deloitte's 2026 work on agentic AI in mid-market finance and operations functions points to the same handful of workflows turning up in business after business. The candidates are usually obvious once someone counts the hours properly. They almost never get counted properly without outside help.
How do I know if our data is ready for AI?
Most isn't, and that's the headline finding of Bain's Technology Report 2026, which puts 90% of leaders saying their data foundations are not fit to scale AI. For a UK mid-market business the gap is usually wider than it looks from inside. The honest answer is that data readiness is not a yes-or-no question. It is a workflow-by-workflow question, and the only way to answer it accurately is to scope each candidate individually.
Should I hire an AI lead or use an agency?
Most UK businesses between £5m and £50m of revenue don't yet have the workflow volume to justify a full-time AI lead. The salary sits north of £90,000 in London for someone credible, and the role takes six months to ramp. An agency engagement is the right shape until the workflows live justify a hire. Once you have five or six agents in production and a roadmap of twenty more, the maths flips and the in-house hire makes sense.
Bottom line for UK founders
Ninety days of well-run AI implementation in a UK mid-market business gets you two or three workflows live, several hundred hours back, and at least one deferred hire. The brands that get there in 2026 reshape their cost base before competitors notice. The ones that don't will be explaining a thirty-person back office in 2028.
Want to see what 90 days could deliver in your business?
The Clerq Diagnostic scopes your specific workflows and produces fixed-price build quotes. From £2,500. 20-minute intro call first.