London has more AI consultancies per square mile than anywhere outside San Francisco. Most of them sell you a strategy deck and a six-month programme. We sell you a fixed quote and a four-week build. The senior judgement stays in London, the team of agents that does the building goes wherever the work is. Here is the honest version of what an AI consultant in London actually does day-to-day in 2026.
We will not name names because the categories shift. But from the conversations buyers have when they shop around, the pattern is clear. London AI consultancies sort into two heaps.
The slideware shops. Offices in Fitzrovia or Shoreditch. A senior partner who used to be at McKinsey, Bain or one of the Big Four. A pitch that opens with "AI strategy" and runs through capability mapping, opportunity sizing, target-state architecture, change-management workshops. After three months and a high five-figure fee, you have a 60-slide deck, a roadmap, and an implementation partner recommendation. You do not have software. The implementation phase is then quoted as a six-month follow-on at six figures.
The delivery shops. A small office or none. The principal is an engineer. The pitch opens with "what is the problem?" and ends with a fixed quote for a specific piece of software in six weeks. After three months you have a working AI system in production. Average project: £8,000 to £40,000, not £200,000.
Both kinds market themselves as "AI consultants in London." The difference shows up in three signals.
Signal 1: what they show you in the first conversation. A delivery shop opens a laptop and shows you something they shipped last month. A slideware shop shows you their methodology and their client logo wall. If the client logos are all FTSE 100 and the work was always with the "innovation team," that is a strategy engagement, not a delivery one.
Signal 2: how they answer "who writes the code?" The honest answer in 2026 is "a team of AI agents, directed by senior engineers." The slideware answer is "we have a delivery partner network we will introduce you to in phase two." If the firm pitching you AI does not itself build software, they are reselling someone else's labour at a markup, and that markup is the reason their project costs three times ours.
Signal 3: what they say when you ask for a fixed quote. A delivery shop will scope, then fix-quote. A slideware shop will resist fixed quotes because they do not actually know what they are going to build. They will sell you a discovery phase, then a strategy phase, then a roadmap phase, then an implementation phase, each scoped only after the previous one. By the time you reach implementation, you have spent six months and six figures, and the software you wanted is still six months away.
These are illustrative shapes drawn from London SME workflow patterns. The fixed-quote bands are the real ranges we quote for builds of this shape; the workflows described are the patterns that fit each context. Almost every project of this kind for a London client begins with the same artefact: an Excel workbook on someone's laptop that the business genuinely depends on. The spreadsheet is named after the person who built it. It has macros nobody else understands. It crashes when more than two people open it. It runs the business.
Profile: a 14-tab Excel tracking customer onboarding KYC status, source-of-funds documents, sanctions screening results, manual override approvals, and audit trail. Updated by two compliance analysts. The CFO could not get a clean "open KYC cases by risk band" view without 90 minutes of pivot-table work each Monday morning. The shape that fits: a KYC workflow application with AI document extraction, risk-banded views and audit-trail capture. Typical band: £12,000 to £25,000 fixed quote, eight to twelve weeks. Expected outcome: the Monday pivot-table session disappears, compliance analysts work in a purpose-built system, audit trail tightens.
Profile: a project pipeline workbook with rows per live brief, columns for client, retainer or project, hours quoted, hours delivered, hours billed, scope-creep flags. Linked to a second workbook for resource allocation across 18 freelancers and staffers. Producers spend six hours a week reconciling the two. The shape that fits: a project pipeline and resource-allocation app, with AI summarising weekly burn and flagging scope creep. Typical band: £8,000 to £15,000 fixed quote, six to ten weeks. Expected outcome: producer reconciliation hours drop materially and scope creep is visible before it bites.
Profile: the live-roles Excel and the candidate-flow Excel and the BD-pipeline Excel, none of which talk to the CRM (Bullhorn) because the CRM is not loved enough to be the source of truth. Senior consultants keep their personal versions, which contradict each other at every Friday meeting. The shape that fits: a pipeline layer over Bullhorn that consultants actually want to use, with AI doing the candidate-status updates and BD-pipeline summaries. Typical band: £10,000 to £18,000 fixed quote, eight to twelve weeks. Expected outcome: a single live view of roles, candidates and BD, replacing the conflicting personal workbooks.
Profile: the matter-pipeline and WIP workbook, rows per matter, columns for stage, fixed fee or time recorded, billed, unbilled WIP, risk flag. The partner's evenings are absorbed by reconciling the workbook against the practice's case management system (Proclaim) because the two never quite agree. The shape that fits: a WIP and matter-pipeline application that sits over Proclaim, with AI summarising matter status and flagging risk. Typical band: £10,000 to £20,000 fixed quote, eight to twelve weeks. Expected outcome: partner evenings recovered; matter risk surfaced earlier.
Profile: a referral-tracking and consultant-fee-share workbook, which underpins both clinical follow-up and revenue distribution. When the practice manager goes on holiday for two weeks, nobody else knows how the formulas work. The shape that fits: a referral-tracking and fee-share application with role-based access, AI summarising open referrals and fee shares per consultant. Typical band: £8,000 to £16,000 fixed quote, six to ten weeks. Expected outcome: referral follow-up and revenue distribution stop depending on one person's formulas.
In every case, the spreadsheet is the symptom. The real problem: a system the business has outgrown but cannot afford to replace at enterprise-software prices. The AI consulting work is the bridge. Six to twelve weeks of fixed-quote build, and the workbook becomes purpose-built software with AI doing the parts a person used to do by hand.
Before you commission anything from us, the honest comparison. Buyers in London have four real options for AI consulting work. Each has a place. The right answer depends on the problem you are trying to solve.
| Option | Typical cost | When it fits | When it does not |
|---|---|---|---|
| Big-four / strategy AI practice (Deloitte, PwC, EY, KPMG, Accenture, Capgemini) | £200k+ programme | Board-level AI strategy, regulatory submissions, FTSE-100 governance work where the partner name on the letter matters | Shipping software for an SME |
| Specialist London AI consultancy (boutique, 10 to 50 staff) | £60k to £300k | Complex multi-system AI builds with regulatory or model-risk overlay, often financial services | A focused workflow automation under £40k |
| Senior fractional AI engineer (this category) | £3k to £40k per project, or £450 to £1,500/mo retainer | Small to mid-market businesses, focused projects, fixed-quote delivery, owned software | Programmes that need 20+ people working in parallel |
| Freelance AI developer (Upwork, Toptal, individual contractor) | £400 to £900/day | A defined task with a clear spec, where you already know what to build | Anything requiring senior architectural judgement or end-to-end delivery accountability |
Many London businesses default to option one when option three fits them better. The default to a big consultancy is a hangover from a world where shipping software was slow. AI agents have changed the unit economics: work that needed a team of 12 in 2020 now needs a senior engineer directing a team of agents.
We will say no to more London AI work than we take on. The patterns that recur in inbound enquiries:
"AI transformation of the agency" (creative, marketing, PR). A founder reads a piece in Campaign or Marketing Week and wants their 20-person Soho agency to become "an AI-native agency." We will tell you that is a positioning project, not a technology project. The right AI work in a creative agency is unglamorous: project pipeline automation, finance reconciliation, briefing-doc QA. The transformation pitch buries those under strategy theatre and never ships them.
"AI sales agent that cold-emails for us." Recruitment agencies, fintechs, and B2B SaaS founders ask for this regularly. The honest answer is that AI cold-outreach tools blow up sender reputation within weeks. Replies are either "please remove me" or worse. The honest path is AI helping your team write better outbound; humans send and reply.
"AI chatbot to replace customer support." If your customer support involves regulated information, distressed customers, or any nuance, a chatbot will damage the relationship faster than it saves you money. The honest version is a small AI assistant that handles the top ten FAQs and routes everything else to a human, fast. That ships in three weeks for £6,000 instead of three months for £60,000.
The shared feature: each is a project where the pitch is exciting and the actual outcome would be a mess. Saying no to those is part of how the consultancy keeps quality high. The good London buyers know to be wary of consultancies that say yes to everything.
Every project is scoped and fixed-quoted before any work starts. Whatever the conversation, you will have the price before you commit. The bands below are the ranges we quote for the shapes of work we take on.
| What it is | Timeline | Fixed quote |
|---|---|---|
| Scoping engagement (a written delivery plan, fixed scope, fixed quote for the build, no commitment to proceed) | 1 to 2 weeks | £1,500 to £4,000 |
| Small focused AI build (one workflow, one system, one team) | 2 to 4 weeks | £3,000 to £8,000 |
| Document processing, invoice automation, AML/KYC automation | 3 to 6 weeks | £5,000 to £15,000 |
| AI assistant over your own documents and policies | 4 to 8 weeks | £6,000 to £15,000 |
| Multi-system AI workflow, custom portal, end-to-end automation | 6 to 12 weeks | £12,000 to £40,000 |
| Tech partnership (we run, maintain, improve your stack) | Monthly, no lock-in | £450 to £1,500 / month |
The strategic point: if a London AI consultancy is quoting you a six-figure number for a first project, you are not buying software, you are buying their cost base. The decks, the partners, the office overhead. None of that builds the thing. The work that genuinely needs a six-figure budget is rare, and we will tell you when your problem is one of those cases (it usually is not).
The natural next step after a successful first project is either a second project, or the tech partnership retainer, where we run, maintain and keep improving the systems we build. Many clients typically move into the retainer within six months of a first build because shipping one piece of software typically surfaces the next one.
I am Oliver Wakefield-Smith. More than twenty years as a developer, tech lead, architect and engineering lead, shipping software for businesses in satellite communications, UK broadcast media, sports media, water utilities, retail, industrial engineering and the public sector, much of it for or with London-headquartered teams. Those companies were not running AI. They were running on the careful integration of systems that did not want to talk to each other.
That experience is now the engine behind a directed team of AI agents. The agents write the code, run the tests, build the integrations, ship the documentation. I direct them, decide what should be built, and stand behind every result. The full delivery shape is on the AI implementation page, and the kinds of custom builds we deliver are on the app builds page.
Adjacent reading: our take on what a good first project looks like, our Power Automate consultant page for the Microsoft-anchored alternative, and our CRM for UK accountants page for an example of vertical-specific work.
Half the AI consultants in London write strategy decks and run workshops. The other half build and ship working software. We are firmly in the second category. The day-to-day work is integrating the systems a London business already runs (Xero, HubSpot, Salesforce, Microsoft 365, your bank feeds, your spreadsheets) and putting AI in the seam where a person was previously moving data by hand. The shapes of project that fit this consultancy in London are document processing, AI assistants over a firm's own content, and automation of repetitive judgement work.
A small focused AI project runs £3,000 to £8,000 as a fixed quote. A larger build, a custom portal or multi-system AI workflow, runs £12,000 to £40,000. A retainer where we run, maintain and keep improving your stack is £450 to £1,500 per month. London day rates from the larger AI consultancies typically sit £1,500 to £3,000 per consultant per day, plus partner uplift, which is why a programme of work runs to six figures fast. Fixed-quote delivery removes that uncertainty.
Yes. The senior judgement, the scoping, the architecture and the standing-behind-the-work all sit in London. The team that does the building is a team of AI agents, which means the work itself goes wherever it needs to. For London clients we can meet in person, on-site at your office, or remote, whichever is most useful for the project.
Ask three questions. First, can you show me a piece of working software you shipped in the last six months? If the answer is a case study about a strategy engagement, that is a slideware firm. Second, who actually writes the code? If the answer is a partner network or an offshore team they will introduce you to later, that is a markup operation. Third, will you fix-quote this? If everything is time-and-materials with a discovery phase that has its own discovery phase, that is a firm that does not know what it is going to build.
The shape that fits is London-anchored SMEs and mid-market: fintech in Canary Wharf, creative agencies in Soho and Shoreditch, recruitment firms in the City and Holborn, law firms in Chancery Lane and Holborn, private medical groups in the Harley Street area, accountancy practices across central London. The common thread is a business with real operations, real data, and a real workflow that AI can take a bite out of.
Yes. The code, the integrations, the prompts, the deployment, all of it. It runs on your own environment. If we part ways, the system keeps running. We do not build on proprietary platforms you cannot leave, and we will not lock you in.
Tell us the problem you are trying to solve. We will tell you honestly whether AI is the right tool, and if it is, scope and quote it before any work starts. No charge for the conversation.
Email oliver@digitalsignet.com