Productized decision tools
The knowledge in one person’s head, as a tool the team opens
Every company has a handful of people who just know what to do next. Everyone else guesses, or asks, or does nothing. This is a working demo of what I build to close that gap — not a slide about it.
This is a demonstration. Harborline Technologies is a made-up managed services provider and the service catalogue below is invented for this page. The logic, the reasoning and the interaction are real — the client is not. Nothing here comes from any engagement.
Harborline Technologies — Next Play
An account manager has just closed something, or inherited an account, and has ninety seconds before the next call to work out what to raise. Three inputs, then a ranked answer with the words to actually say.
Step 1 — Anchor
What did you just sell them, or what are they already running with you?
Step 2 — Already covered
Anything else they have, from you or anyone else. Optional — but it is what stops you pitching something they bought last year.
Step 3 — What’s driving the conversation
Optional. Changes the order, not the reasoning.
What is actually going on underneath
It looks like a lookup table. It is not, and the difference is the whole point.
The reasoning is the product
Anyone can list adjacent services. The value is in why now — the mechanism that makes this the right next conversation for this account, in language a new hire can repeat without sounding like a brochure.
It subtracts as well as adds
Telling someone what to pitch is easy. Telling them what to stop pitching is what earns trust in the tool. Anything already in place drops out and is shown as such, so nobody has to remember.
Every result shows its work
Each recommendation displays how it scored and why it moved. A tool that produces a number nobody can interrogate gets used twice and then abandoned.
It ends in an action
A discovery question and a thing to verify first. A recommendation with no next sentence attached is a suggestion, and suggestions do not change behaviour.
The content is editable by the business
The map lives in one structured block, separate from the interface. The person who owns the offering edits it — not a developer, and not me on a retainer.
One file, no infrastructure
This whole page is a single static file with no backend, no login and no database. It loads on a phone in a parking lot, which is where it will mostly be used.
What yours would look like
Same shape, different content: a cross-sell recommender over your catalogue, an onboarding readiness check, an eligibility or qualification screen, a pricing configurator with the guardrails built in. The work is roughly a third interface and two thirds sitting with the people who already know the answers and getting it out of them in a form that survives their resignation.
Common questions
Is this AI?
This demo is not, deliberately. The logic is explicit and auditable, which for a rules-driven decision like this is a feature — the business owns the reasoning and can change it without retraining anything. Where a problem genuinely needs judgement over unstructured input, I use a model for that step and keep the deterministic parts deterministic. The failure mode I care about is a tool that confidently produces a plausible answer nobody can check.
Why not just build this in our CRM?
Often you should, and I will tell you so. In-CRM makes sense when the recommendation needs live account data and the team already lives in the CRM. A standalone tool wins when you need it in front of people who do not have licences, when you want it usable before a renewal cycle rather than after, or when the CRM changes are queued behind six months of other work.
Who maintains the content?
You do. The map is a separate, plainly structured block with one entry per relationship. I hand over a short guide for editing it and I would rather you never call me about a wording change. If the underlying offering changes shape entirely, that is a different conversation.
What do you need from us to build one?
Access to the two or three people who already make these calls well, for a few hours. That is the real input. Everything else — catalogue, pricing rules, the constraints nobody wrote down — comes out of those sessions. If those people cannot be freed up, the project will not work and I would rather say so before you pay me.
Can we see something closer to our own business?
Yes. A rough version in your own domain, with your real catalogue, is usually a day or two of work and is a much better basis for deciding than this page is.
Contact
Tell me what’s stuck.
A short call is free and usually enough for both of us to know whether there’s a fit. Bring the messy version — I don’t need it cleaned up first.