R&D operations

Why a grant consultant is not enough for an AI R&D project

A grant consultant knows how to write a Business Finland application. That is a real skill and it clears a real hurdle. But for an AI R&D project, the application is the second-order problem — the first-order problem is whether the project itself is shaped in a way that is genuinely fundable and buildable. A grant consultant who is not also an AI operator cannot shape that project. They can only polish whatever plan you bring them.

Grant consultants are useful. On a normal Business Finland R&D application, a good one saves weeks of work, brings hard-won knowledge of what evaluators want to see, and turns a rough plan into a submission that clears the first read. Nothing in this piece is an argument against hiring one.

The argument is more specific: on an AI R&D project, a grant consultant working alone is not enough. The reason is structural, not personal. Grant consulting is a proposal-writing craft. AI R&D is a technical shaping problem. The two require different senior people, and pretending one person covers both is where funded AI projects go wrong.

What a grant consultant is genuinely good at

Before the critique, the actual value. A skilled Finnish grant consultant:

Knows the current Business Finland instruments, their eligibility windows, and their unwritten scoring habits
Turns a fuzzy company narrative into a structured application with work packages, budgets, and a defensible risk section
Writes in the register evaluators expect — precise, evidence-grounded, unhyped
Handles the mechanical machinery: budget logic, cost categories, de minimis calculations, consortium terms if relevant
Runs the schedule so the submission actually happens on time

This is real work. It is skilled work. A company trying to write its first Business Finland application without help usually spends three to four times longer and produces something visibly weaker. Nothing here disputes that.

What a grant consultant is not, on an AI project

The line to be honest about: a grant consultant is a proposal writer, not an AI operator. Those are different jobs, and an AI R&D application asks the second one much more than the first.

Concretely, a proposal writer working on an AI R&D project cannot, on their own:

Judge whether the proposed model architecture is the right choice for the data you actually have
Push back on a CTO who is over-scoping or under-scoping the ambition of the project
Distinguish research-grade AI R&D — where genuine uncertainty exists — from an implementation project dressed up in R&D language, which evaluators increasingly detect
Evaluate vendor claims and vendor risk on the technical side
Argue for or against specific technical choices with a technical evaluator during the review
Anticipate which reporting evidence Business Finland will actually want when the project runs

None of this is a criticism of grant consultants as people. It is a description of the job. The proposal writer’s craft is the shape and rhetoric of the application. The AI operator’s craft is the substance of the project. On a plain innovation grant, the two can be pried apart. On an AI R&D grant, they cannot.

Where the mismatch shows up

The pattern is consistent enough to be listed. Companies using a grant consultant alone on an AI R&D project tend to see one or more of the following:

The company writes the AI project itself, then pays for polish. The consultant is skilled at prose but cannot argue with the CTO about technical scope, so the CTO ends up writing the substantive plan and paying the consultant mostly for framing.
The grant is smaller than it should have been. The project is scoped conservatively because the consultant cannot confidently argue for the more ambitious version. Ambition costs money to defend; consultants who cannot defend it hedge downward.
The application overpromises in ways the evaluator can see. In the opposite failure mode, the consultant pushes ambition upward for narrative reasons, and the technical feasibility claims stop matching reality. Evaluators are increasingly good at spotting this.
The technical risk section is generic. Real AI risks — data quality, model failure modes, integration cost, generalisation — get flattened into generic R&D risk language. Evaluators notice.
The kickoff is a re-planning exercise. Once the money lands, the delivery team quietly rewrites the project because the plan submitted was not the plan they can execute.
Reporting drifts from the plan. The consultant is often no longer engaged, and nobody remembers exactly which milestones were framed to which evaluators. Reporting becomes a translation exercise.
Mid-project scope changes are hard. Any legitimate R&D scope change has to be renegotiated with Business Finland, and the person who framed the original plan is no longer in the room.

Each of these has one root: the shaping of an AI R&D project is technical work, and the person shaping it needs to be an AI operator, not only a proposal writer.

The specific test: shape drives grant size

There is a useful diagnostic. Ask a grant consultant, before hiring them for an AI R&D project:

“If we told you we want to apply for a EUR 300k R&D grant, can you help us understand whether the project we have is more naturally a EUR 150k project, a EUR 300k project, or a EUR 700k project — and change the shape if the honest answer is one of the other two?”

A grant consultant working alone will usually say something reasonable but non-committal: “we can size the budget to the ambition you describe.” That is honest. It also means the sizing decision — which is where most of the grant’s value is decided — is still yours to make alone.

An AI operator will answer with technical questions about the project itself: what data, what model class, what integration, what learning to be earned, what would be feasible in 18 months. Those questions are what determine whether the project is a EUR 150k, EUR 300k, or EUR 700k project. Shape drives grant size.

Grant consultant vs AI operator: side by side

	Grant consultant	AI operator	Both, on the same engagement
Core craft	Proposal writing	AI project shaping and delivery governance	Proposal that is fundable and buildable
Judges technical scope	Rarely	Yes	Yes
Argues with the CTO	Not effectively	Yes	Yes
Knows Business Finland instruments	Yes, in depth	Sometimes	Yes
Writes the application	Yes	Sometimes	Yes
Sits in delivery	Rarely	Yes	Yes
Handles technical reporting evidence	Partially	Yes	Yes
Best when	The project is straightforward, non-AI R&D, and you need submission help	You already have strong AI leadership and need shaping and governance	The project is AI R&D, senior AI leadership is thin, and grant success matters

The right column is not “hire two vendors.” It is “hire one operator whose craft covers both.”

Two honest exceptions

There are two shapes of AI R&D project where a grant consultant alone is sufficient, and it is worth naming them:

The company already has a strong internal senior AI lead who can shape the project independently and defend it in the review. The consultant then does what they do best — turn a well-shaped plan into a well-written application. This is fine, and it is roughly the model most grant-writing shops optimise for.
The “AI R&D” project is only lightly AI — most of the technical uncertainty is elsewhere (a novel material, a mechanical design, a biotech step), and the AI component is a supporting element rather than the load-bearing R&D. Here a proposal writer can compose the whole plan without deep AI shaping, because AI is not what the grant is really funding.

Outside those two cases, a grant consultant working alone on an AI R&D project is doing a partial job. That is not a criticism — it is a description of scope.

Consultants write the application and leave; agencies build the tool and leave; an operator shapes the project, defends it, and stays through delivery.

What the operator model changes in practice

On BRNSFT Capital’s engagements, the shaping-plus-application-plus-delivery arc has produced four approved Business Finland R&D grants totalling EUR 1.57M across 2021–2026:

EUR 70k (2021)
EUR 957k (2024)
EUR 187k (2025)
EUR 357k (2026)

The interesting number is 957k. That grant did not exist as a EUR 957k project when the conversation started — the company arrived with a rough R&D idea that would have been submitted, if written by a consultant alone, as a much smaller project. The size grew because the shape of the AI R&D changed: broader research ambition, clearer feasibility framing, a technical scope defensible against a technical evaluator. Shape first, then application, then delivery. That is the operator model working across the arc, and it is what a proposal writer working alone structurally cannot deliver.

FAQ

Should I hire a grant consultant for an AI R&D project? Yes — if the project is well-shaped already and you need submission craft, budget mechanics, and evaluator-facing prose. That is what grant consultants are for and what they are good at. What they cannot do alone is shape the AI R&D itself. If shaping is still needed, add an AI operator to the engagement or work with someone who covers both crafts.

Isn’t it cheaper to use just a grant consultant? On the consultant’s fee, yes. On the total cost, usually not. A conservatively shaped project is a smaller grant. A generically framed technical risk section reduces the odds of approval. A weak plan makes delivery more expensive. The saved consulting fee tends to be re-paid, and then some, downstream.

How do I tell if my project actually needs an AI operator, not just a grant consultant? Two signs. First: your internal team cannot yet write the substantive AI project plan on their own — the technical scope, model choice, and data strategy are not yet resolved. Second: the ambition and risk framing of the project would change materially depending on those unresolved decisions. If either sign is true, the shaping work is unfinished, and only an AI operator can finish it.

Aren’t some Finnish grant consultancies now specialising in AI? Some claim AI-specific expertise, and a small number genuinely have senior AI operators on staff. The honest test is not the marketing; it is whether the person actually working on your project can push back on a CTO about model choice, data readiness, and technical scope. If the answer is “our AI specialist reviews the application after our writer drafts it,” that is proposal writing with an AI review layer — closer to the model this piece critiques than to a real operator model.

What about Business Finland’s own advisory service? Business Finland’s own advisors are genuinely useful — they will discuss instrument fit and give directional feedback on early-stage ideas. What they cannot do is shape your specific AI R&D project or write your application. They sit above the process, not inside it.

If we already have a CTO, can our CTO shape the AI project themselves? Sometimes, yes. Two things to check honestly: whether the CTO has the specific senior AI shaping experience (framing an R&D project against a technical evaluator is a different skill from running an engineering team), and whether they have the calendar time to do the work alongside their existing job. If both are true, you don’t need an external operator for shaping — you need a grant consultant for writing.

The one-sentence version

A grant consultant polishes the application you bring them, an AI operator shapes the project itself — and on an AI R&D grant, the difference between the two is where the grant’s actual value lives.