AI consulting should end in something that runs
A good AI consultant reads your real workflows and data first, and the engagement ends in a buildable plan or running software, not a deck.
Key takeaways
5 things that decide this
- 01A good AI consultant reads your real workflows and data before recommending anything, and the engagement ends in a plan an engineer can build from, or in software that runs, not in a slide deck.
- 02The diagnosis is the valuable part. On TrialTriage, reading how nurses actually worked showed that the search, not the eligibility decision, was eating the hours. That changed what got built.
- 03Sometimes the fix is replacing a tool. Go4Gr8 had outgrown TypingMind Custom, which needed more than two hours of manual setup per user.
- 04Our Blueprint is a written plan you own. You can take it to another firm.
- 05The trade-off: a consultant who also builds has an interest in the build. A fixed fee and a plan you can take elsewhere are how that interest stays honest.
What are the two kinds of AI consulting?
Most AI consulting falls into one of two kinds, and you can usually tell which one you are buying from the final deliverable.
One kind ends in a deck. It covers the market, the opportunity, a maturity model and a roadmap with phases. It may be well researched. But no engineer can build from it, because it never touched your real data or the way your team actually works. After the deck, the next step is usually another engagement to work out what the deck meant.
Another kind ends in something that runs, or in a plan specific enough that an engineer could start on Monday. To get there, the consultant has to read your workflows and your data before recommending anything. That reading is the work. It is also where most of the surprises are.
If you are hiring an AI consultant, ask to see a final deliverable from a past engagement. If it is a deck, ask what was built from it.
How did the diagnosis change TrialTriage?
TrialTriage helps oncology nurses match patients to clinical trials. The obvious brief was 'use AI to decide which patients are eligible for which trials'. That is the part that sounds like AI.
But the diagnosis said something else. We mapped how nurses manually cross-checked patient records against NCCN guidelines and trial databases. In the words of the case study: 'The search, not the eligibility decision, was eating the hours.'
That changed the design. The platform uses large language models to rank matching trials against NCCN guidelines and drug data, which is the search. A nurse reviews, adjusts and finalizes every result, which is the decision. Clinical judgment stays with the nurse, and the hours come out of the part that was actually slow.
A deck written without that reading would have aimed the AI at the decision. That is harder, riskier and not where the time went.
When is the answer to replace a tool?
Not every diagnosis ends in a new AI feature. Sometimes it ends in 'the tool you have cannot do this'.
Go4Gr8 gives executives AI sparring partners for hard decisions. They had built on TypingMind Custom. It had no multi-organization model, so a B2B product could not run on it. Every new user and agent took more than two hours to set up by hand.
No prompt change fixes that. The fix was a custom multi-tenant platform where each organization manages its own users and invites, and every person gets dedicated AI agents. Go4Gr8's case study reports that a new user is active in under 15 minutes. That is Go4Gr8's onboarding measure for their platform, not a promise about yours.
The lesson for anyone buying AI consulting: a good consultant will tell you when the problem is the plumbing, not the model.
What does a Blueprint hand you?
At Hashlogics, every engagement starts with a 45-minute call and then a fixed-fee Blueprint. An engineer reads your real code, data and workflows before we quote. The Blueprint is the written result.
It names the problem as the data shows it, not as the brief guessed it. It says what to build, what to leave alone, and which systems it touches. It sets out the build in milestones, each with written acceptance criteria. It also lists the risks we found, including the ones that argue against building.
You own the Blueprint. You can take it to another firm and have them build from it. If the plan isn't one you'd act on, you don't pay for it. If you go ahead with us, the fee is credited in full against the build. Read more on our AI consulting page.
Should an AI consultant also build the system?
There is a fair objection here. A consultant who also builds has an interest in the build. They may find a reason to build something you do not need.
We guard against that in three ways. The Blueprint is a fixed fee, so its price does not grow with the size of the build it recommends. It is yours to take elsewhere, so you are never locked in. And it states the case against building where there is one.
The upside of a builder-consultant is that the plan has to survive contact with the engineers who wrote it. A plan written by people who will not build it never faces that test.
What should you ask an AI consultant before you hire?
Will you read our real data and workflows before you recommend anything? What is the final deliverable, and can an engineer build from it? Can we see one from a past engagement? Who owns it, and can we take it to another firm? What did you recommend against building last time?
If the answers are vague, the deliverable will be too.
If you want to see how we answer them, the next step is a 45-minute call, then a fixed-fee Blueprint written from your real systems, credited in full against the build. See how we work.
FAQ
01What does an AI consultant actually do?+
A good one reads your workflows and data, finds where the time or money actually goes, and recommends what to build or change. The engagement should end in a plan an engineer can build from, or in software that runs. A deck alone is not enough.
02How do I choose an AI consulting firm?+
Ask to see a past final deliverable and what was built from it. Check that the firm reads your real data before it recommends anything. Ask who owns the plan and whether you can take it to another firm. Ask what it recommended against building.
03Should an AI consultant also build the system?+
It can work well if the plan is fixed-fee and yours to take elsewhere. A builder's plan has to survive the engineers who will build it. The risk is a bias toward building, so ask how the firm guards against it.
