Ares Intelligence Raises £2.48M Seed to Map AI Use Inside Firms

Ares Intelligence has raised about £2.48m in seed funding to tell companies where AI would actually save them money.

Creandum and Lakestar led the round, joined by unnamed angel investors with Google DeepMind ties. The raise was announced on 30 October 2026.

The London company was founded in 2026 by Matt Falconer and Alfie Ranstead. It runs an AI interviewer that holds confidential voice conversations with employees in more than 70 languages, then turns what it hears into a ranked list of where AI would cut time and cost.

Leadership gets that list within weeks rather than after a consulting engagement.

Reported figures for the round differ. Dealroom puts it at £2.48m, which it describes as roughly €3m, while other outlets have reported €2.6m. The company has not published a figure of its own.

What the investors actually bought

Creandum and Lakestar are not small cheques looking for an AI label. Creandum also co-led the record £40m UK robotics seed round for Machenta earlier in October, and Lakestar has been an active European early-stage name for well over a decade.

What they backed here is not a model or a tool. It is a diagnosis service.

Ares says it has enterprise customers across Europe and the United States inside roughly seven months of existing. At Tremonia, a Mercedes-Benz spin-out, it reports finding more than 400 hours of avoidable manual work per month within four weeks.

That is the pitch in one line. Not "here is an AI", but "here is where yours should go".

The pivot that explains the product

Falconer and Ranstead did not start here. Ares grew out of their previous venture, Instruct, which built automation. Customers wanted to automate and could not tell them what to automate first.

That detail is worth more than the funding number.

For three years the constraint on enterprise AI was capability. Models could not do the work reliably enough to trust. That constraint has largely gone, and a different one has replaced it: nobody inside the building can say with any confidence which processes are worth touching.

The MIT NANDA research widely cited through 2025 and 2026 found that around 95% of corporate generative AI pilots returned nothing measurable. A European Central Bank survey in March put implementation difficulty and skills shortages at the top of the barriers firms report.

Both findings describe the same failure. Companies bought capability and then pointed it at the wrong things.

Ares is selling the pointing. So is Mendo in Paris, which raised €12m in June for a similar job and has taken about $33m in total. That is a far larger war chest aimed at the same gap, which tells you the gap is real and that Ares is not early to it.

The part worth thinking about

An AI system that voice-interviews your staff about their own inefficiencies, then hands a roadmap to the people who decide headcount, is workplace monitoring whatever else it is.

Ares has clearly thought about this. It says the platform is UK GDPR and SOC 2 aligned, and that it can mask names, redact voices and strip role metadata before anything reaches leadership.

Read that list again. Every item on it is a defence against the obvious objection, which is that honest answers get people managed out.

The ICO's guidance on data protection and monitoring workers is the relevant standard for any UK deployment, and it turns on whether workers would reasonably expect the monitoring and whether a less intrusive route exists. Anonymisation helps with the data protection question. It does not fully answer the trust one, because employees still know the recording is happening and still choose what to say.

None of that will sink the company. But it decides whether the interviews produce anything true, and true answers are the whole product.

What this means for UK founders

The useful lesson is the pivot, not the round.

Falconer and Ranstead had a working automation business and a customer problem sitting one step upstream of it. They moved to the upstream problem. Most founders in that position try harder to sell the thing they already built.

Second, the compliance architecture is the feature here, not overhead bolted on for enterprise procurement. Masking and redaction are what make the raw material usable. UK founders selling into regulated enterprises should read that as a pattern rather than a cost.

Third, a £2.48m seed is a modest entry into a category where a Paris competitor already holds about $33m. Our breakdown of which UK cities and sectors take the most startup funding shows how concentrated UK rounds have become, and being well-capitalised relative to your market matters more than the absolute number.

The adoption gap Ares is selling into is not going to close on its own. Anthropic's own data showing AI agents running six hours for every human hour suggests capability keeps moving while most organisations stand still. Somebody has to work out where to point it, and right now that job is wide open.