Become an AI-native company.
If you treat AI as a tool you bolt on, you'll get a faster version of the old company.
But build it as the operating system you run on, and a small team can outpace much bigger rivals.
Book a call
What it means to be AI-native
An AI-native company isn't one with more AI tools. It's built on three things.
A company brain
Everything your company knows, captured so AI can actually use it.
AI at the center
AI runs routine operations; people do the work that needs a human.
Systems that self-improve
They watch their own results, catch what's not working, and get better on their own.
Who you're working with
Experimental Technology was founded by Huw Walters. Over the past year we've worked hands-on with small and mid-sized businesses putting AI at the center of how they operate. Huw has worked in AI since 2018, including as a Staff AI Product Manager at Meta and as Head of Product at an AI startup acquired by TikTok, and holds a degree in philosophy and psychology from the University of Oxford.
We're not a platform or a big consultancy. We're a small, senior team that works alongside yours, blending advice, engineering, and hands-on teaching so your people can run what we build themselves.
Things we've built
A regulation brain over a vast legal corpus.
We built a system that reasons over a vast corpus of scanned legal and policy documents. An expert asks a question in plain English and gets an answer grounded only in that material, drawn from real source passages. A custom interface lets them inspect each passage behind an answer, highlighted against the original document, so a huge body of material becomes something they can explore and trust rather than wade through.
A company brain for an engineering team.
We built a living, queryable knowledge layer over an engineering team's projects, decisions, and tooling: a single trustworthy place to ask what the status of something is or why a choice was made. And because the full context sits in one place, AI can draw on it to write highly contextualised product requirements, tests, and documentation.
Turning sales calls into usable insight.
We built analysis over sales call transcripts that maps what was said to a structured value-and-pain framework: which pains came up, who raised them, and how strongly. It aggregates recurring themes across many calls, so commercial teams act on consistent, evidenced signal rather than memory or one-off summaries.
A marketing engine for fast, on-brand campaigns.
We gave a commercial team a shared library of AI skills as a Claude Cowork plugin, so anyone can produce on-brand, consistent assets from the same building blocks, then wrapped it in a Notion front end wired to Claude routines. An operator inputs their requirements on a card and kicks off generation; an agent produces the full set of campaign assets, a person reviews them, and approved assets publish out to the website and to social.
An AI researcher that feeds the company brain.
We built a research agent that works across the open web and other sources, grounded in a company's own knowledge base, to produce briefings good enough to inform real decisions and to seed published writing. What it learns feeds back into the company brain, so research compounds into shared knowledge rather than producing one-off answers.
An AI data scientist for teams without one.
We built a tool that works like a data scientist on call. It helps you brainstorm which questions are worth asking of your data given your goals, then turns your plain-language questions into the underlying analysis and returns the answer, sometimes as a direct figure, sometimes as a clear visualisation.
Your questions,
answered
Much of what makes a company work isn't written down. It's tacit expertise held by a few key people. We use structured interviews, voice agents, and targeted capture sessions to surface this knowledge and fold it into your company brain. This is especially valuable where the current documentation has gaps.
Messiness is what your actual organization runs on, and the company brain is designed for it. We weigh sources by recency, type, and relevance, surface conflicting information rather than silently merging it, and route ambiguity to the right subject-matter expert when it matters. Drafts, emails, transcripts, scanned documents and mixed-quality material are all in scope.
We design the AI we build to recognize when information is incomplete, ambiguous, or based on assumptions. Rather than fill gaps with confident-sounding guesses, the outputs highlight uncertainty, show partial evidence, and flag areas for human review.
Security is designed in from day one of every engagement. We work with you to understand your data sensitivity, governance needs, and risk profile, then design an architecture that fits. That ranges from secure cloud deployments with strict access controls to private or local model deployments where data must not leave your environment. Your data is never used to train public models.
No, not in the small and mid-sized companies we work with. Larger organizations have traditionally carried more headcount than the work strictly needs, which is where the shrinking is happening. The companies we work with have the opposite problem: more work than their people can get through. With AI at the center, a small team can operate at a much larger scale.
We agree what "working" means with you at the start of each engagement, then build the evaluation in so you can see it for yourself. Scoped early engagements (a first company brain, an audit of existing tools) typically produce a usable result in weeks, not quarters.
Yes. A common first engagement is reviewing AI tools you've already invested in against your real use cases and standards. You walk away knowing where to trust them, where they fall short, and whether the gaps are worth fixing.
Most relationships start with one scoped piece, often a first company brain build or a review of AI you've already bought. From there they usually grow into an ongoing partnership as the systems and the trust compound. We're a fit if you want to start narrow and see if it works before going wider.
We work with ambitious small and mid-sized companies, typically 20 to 200 people, where leadership engages directly with AI strategy. The fractional, embedded model fits those environments. The Big-4 playbook (large teams, twelve-month timelines, six-figure minimums) doesn't, and we don't try to run it.
Get in touch
Ready to explore what becoming an AI-native company could look like for you? Let's start a conversation.