AI consultancy in London that ends in a decision, not a slide deck.

We work with owner-managed and mid-sized businesses in London and across the UK: recruitment agencies, care providers, accountancy practices, law firms, clinics, agencies, trades. Typically ten to two hundred staff, a process that has outgrown the spreadsheet, and a director who has been quoted for AI and cannot tell whether the number is reasonable.

Be honest with yourself before you hire anyone, us included. If you have fewer than ten staff and no repetitive, rule-based work that costs real hours, a consultancy engagement will not pay back. Give the team a paid ChatGPT or Claude subscription, agree three rules about what may not be pasted in, and see what they do with it for a month. That advice is free and it is the right advice for a lot of businesses.

Consultancy earns its fee in three situations. When people spend hours a week reading, sorting or retyping information that arrives in a predictable shape. When systems that should talk to each other do not, and a person is the integration. And when you handle data you must not feed into a public model and need to know where the line is.

Interviews with the people doing the work, not only the people describing it. An audit of the data you already hold and the state it is in. A map of the opportunities ranked by return rather than by novelty. A cost model for each one, including what the AI would cost to run per month once it is live, which is the figure almost every proposal leaves out. And a roadmap from first use case to production, with prices and timelines attached.

The output is a written document, specific enough to build from and yours to take to any supplier, including one that is not us. It is not a presentation.

The honest market picture, so you can judge any quote including ours. Published UK ranges for AI consultancy cluster into an audit or readiness sprint at roughly £2,000 to £5,000, advisory day rates of £500 to £1,500 depending on the firm, and full builds from £10,000 upward. The large consultancies rarely engage below £25,000 and prefer rolling retainers. Boutique firms in London typically quote £200 to £500 an hour or fixed-price pilots.

Our prices sit inside those bands and are fixed in writing before work starts, with an explicit list of what is not included. We do not work on day rates: a day rate pays a supplier for being slow and penalises them for being fast.

Market figures are published UK ranges gathered in September 2026; they are context, not a quote.

Week one: a 30-minute call to hear the problem, then interviews with the people who do the work, then the data audit. You see the opportunity map before the week is out.

Week two: the ROI model, the cost-to-run model, the roadmap and the written report. A walkthrough with the decision-makers, and a plain recommendation, including the option of doing nothing yet.

If there is a build worth doing, the audit fee is credited in full against it. If there is not, you have paid for a document that will save you from the wrong build, which is the cheaper of the two outcomes.

We build and run two AI products of our own. Onrolo screens candidates by voice and chat, chases applicants by message and books interviews without a recruiter awake. Gorizzume tracks more than 120,000 licensed sponsors, refreshed daily, and writes and scores documents for job seekers. Every model call in both is attributed to a feature and a customer, so when we tell you what an AI feature will cost to run, it is a number we have watched on our own invoices, not an estimate.

That record has taught us things a strategy deck cannot. The model is rarely the expensive part of an AI product; the channels around it are. The small model does most of the work, and routing one high-volume feature from the frontier model to the small one made it several times cheaper and faster with no drop in success. And roughly one AI outbound call in eight reaches a human, which is why we design the fallback before we design the voice.

There are four ways to buy AI advice in London. They are not interchangeable, and the right one depends on your size more than on the quality of the firm.

Most businesses should start with the tools they can buy: a ChatGPT Team or Claude workspace, Microsoft Copilot if you live in Microsoft 365. We help teams deploy those properly, with a written policy on what may go in, and we run training sessions on them. That is often the whole answer for a firm under twenty people.

Custom becomes the right answer when the work has to happen unattended, at volume, inside your own systems: reading every application that arrives, answering the phone at 9pm, chasing every invoice, reconciling three systems that were never meant to talk. That is where we build, and where the audit tells you whether it is worth building.

Augustova is registered with the ICO (ZC152144) and a data processing agreement is available on request. Every recommendation in the audit states which data would be processed, by which provider, in which region, and under what lawful basis. Where a use case touches special-category data or automated decisions about individuals, we say so and describe the safeguards, or advise against it.

If your project qualifies, HMRC R&D tax relief can cover a share of the qualifying development cost. We are not tax advisers, but we will structure the scope so your accountant can make the claim.