About
Working systems and confident people.
Fundly AI Consulting exists because the hard part of AI was never the model. It's encoding what your experts know, connecting it to the systems the work lives in, making it dependable enough to trust — and making sure people actually use it.
The position
Two problems, usually treated separately.
Most organisations attacking AI end up with one of two half-solutions.
The first is technology without adoption: an impressive internal tool that four people use, because nobody's job was to teach the other four hundred what it's for or when to trust it.
The second is enthusiasm without infrastructure: everyone's been trained, everyone's keen, and everyone is still copying and pasting between six browser tabs because nothing was ever connected to anything.
Neither half works alone. We do both deliberately — the engineering and the enablement — because in practice they're the same problem seen from two directions: getting knowledge out of individual heads and into systems people can rely on.
What that looks like in practice
- An operations lead's month-end procedure becomes a skill the whole finance team runs.
- A CRM that the AI can actually read, with permissions enforced properly rather than requested politely.
- A prototype that becomes a service with tests, evals, monitoring, and a cost ceiling.
- A department that stops asking "what do I even use this for?" and starts asking better questions.
Any one of those is a decent quarter. Together they're a different organisation.
How we work
Six commitments we'd rather be held to
These aren't values-page filler. Each one costs us something specific, which is the only reason they're worth stating.
We'll tell you not to build it
A meaningful share of "AI problems" are a missing report, an unindexed database, or a badly designed form. Recommending the two-hour fix over the two-month project costs us revenue every time we do it. We do it anyway, because the alternative is a client who eventually works out we didn't.
Vendor-neutral, no exceptions
No referral fees, no reseller margins, no partnership tiers influencing what we recommend. If the built-in feature in software you already own solves it, that's the recommendation — even when it means no engagement.
Handover is the deliverable
Source in your repository, docs your team can follow, a runbook for when it breaks, and a working session with whoever inherits it. No proprietary layer, no hosted component you have to keep renting from us.
Small scope, shipped
We'd rather ship one narrow thing in three weeks than deliver a strategy document in three months. Real usage teaches you things no discovery phase will, and it does it before the budget is spent.
Measured, not asserted
Evals before optimisation. Cost dashboards before cost claims. Before/after benchmarks rather than adjectives. If we can't show you the number, we'll say we're guessing.
Plain language, always
With engineers and with the finance team alike. If an explanation needs jargon to survive, it's usually hiding something. The coaching practice exists because most AI communication fails this test badly.
Fit
Who this works well for
- Small and mid-sized organisations without an internal AI team — where one good engagement genuinely changes how a department operates.
- Teams with a stalled pilot that need the last 90% built rather than another proof of concept.
- Non-technical departments — finance, operations, legal, HR, marketing — sitting on repetitive, document-shaped work.
- Engineering teams who want AI features done properly and would rather pair than outsource.
- Leaders who need a straight, non-hype read on what's feasible this quarter before committing budget.
And who it doesn't
- Anyone wanting an AI strategy deck with no intention of building anything. There are firms that do that well; we aren't one.
- Projects where "use AI" is the requirement and the problem hasn't been identified yet. We'll help you find the problem first, but we won't skip that step.
- Fully autonomous agents with production write access and no human in the loop. We'll build the agent; we'll insist on the approval gates.
- Work that depends on overstating what AI can do to an internal audience or to customers.
Process
What working together looks like
1 · Scope
A free 30-minute call, then a written recommendation with an effort estimate — including "don't do this" where that's the honest answer. Yours to keep either way.
2 · Agree
A written scope defining what "delivered" means, a fixed price where the work is knowable, and a start date. No ambiguity about what you're buying.
3 · Build
Weekly written updates, working software you can see rather than status reports, and scope changes flagged the moment they appear rather than at invoice time.
4 · Hand over
Code, documentation, runbooks, and a live session with the people who'll own it. Two weeks of post-launch support, then you're genuinely independent.
Data & security
How we handle your information
Stated plainly, because it's usually buried:
- Least data that does the job. We ask for the narrowest access that makes the work possible, and give it back when the engagement ends.
- Your environment where possible. Development happens against your systems rather than copies on our machines whenever that's practical.
- No production data moved without written approval. Ever, and not as a formality.
- Secrets stay in secret stores. Never in prompts, client code, or conversation history — all three persist and are readable later.
- Documented data flow. Every engagement that touches a third-party model gets a written record of what is sent where, so you can answer that question without asking us.
- Your constraints shape the architecture. Zero-retention requirements, regional processing, self-hosted models, air-gapped deployment — all workable, but they're design decisions, so they belong in the first conversation rather than the security review.
We're happy to sign a DPA or NDA and to complete standard vendor security questionnaires. If you have a template, bring it to the scoping call.
Start with one workflow.
The best first conversation is a specific one. Bring the task your team complains about most.