Constantine Consulting
Most firms install the tool and leave. We start with a three-month engagement, and we build it so that by the end of it you can decide what happens next from evidence rather than from a brochure — including deciding you don't need us. Five phases. Click through them.
Why three months, and why a decision at the end of it
Put in a CRM and you have a CRM. It sits there. It does the job you bought it for, and in a year it is doing the same job.
AI does not behave that way. Within a month of giving capable people a capable assistant, two of them are building things nobody approved, two departments are solving the same problem separately, spend is moving in a direction nobody is watching, and your leadership team cannot see any of it. That is not failure — it is what early success looks like, and almost nobody plans for it.
So three months is not a trial. It is the length of time it takes to set the rules, configure the thing properly, train the people who matter, and find out which of them can carry this without us. At the end you know something you cannot know now: who your champions actually are.
The path
Claude's technical settings encode operating choices. We make those choices deliberately — with leadership — before anyone gets a login.
How the company actually works today.
How leadership wants AI to operate.
The governing decisions and technical specifications we will build from.
Applies to every employee, contractor and temporary worker with a Northvale account. Reviewed by the leadership team each quarter.
Claude, on the company Enterprise account, is the approved assistant for company work. Personal accounts and free tiers are not to be used for anything containing customer, employee or job data — including drafts.
Customer bank or card details. Employee medical, disciplinary or payroll records. Contracts before legal review. Anything covered by an NDA that names a third party. Site photographs showing a client's security arrangements.
Bids, proposals, change orders and any letter leaving the company are reviewed by a named person before they go. AI drafts them; it does not send them.
Anyone may prototype on their own time and their own data. Nothing touches live job costing, connects to another system, or reaches a customer until it has gone through the idea queue and been approved.
New connectors, new data sources, and anything that can write into another system: the owner and the IT manager together. Everything else: your division head.
A written AI operating policy your leadership has approved, and a deployment blueprint covering platform, permissions, connected tools, skills, workspaces and training priorities. Phase two turns that blueprint into your configured instance.
This is the step almost every rollout skips. A powerful tool is handed over on factory settings with a training link and a good luck — and the first experience everybody has is a blank box that knows nothing about the company they work for.
Day one looks like your company, not a blank box. Nobody's first hour is spent wondering what to type, and nobody forms the opinion that this thing is not for them.
This is not a company-wide census. You tell us who is key to this — the people whose jobs change most, or who you need carrying it afterwards — and those are the people we assess and train. Two scores each: how willing, and what they can actually do.
The org chart nobody has: who will run with this, who will break something, and who is quietly ready. Almost every leader we show it to finds at least one person they had completely wrong.
Generic AI training produces people who can describe what the tool is. Role-based training produces people who used it that afternoon. We do the second — and we harvest what it generates while we are in the room.
Every person can do their own job with it. Not an AI class — their job. And by month two you are holding a list of everything your company wants built, written by the people closest to the work.
Most consulting engagements end when the invoice does. This one ends at a decision you are equipped to make — because you now have evidence about your own people that did not exist twelve weeks ago.
Drag to reorder priority.
Three months in, you are not guessing. You know who your champions are, what your people can do that they could not in week one, and what your company actually wants built. Handing the reins to your own person becomes a real option rather than a hope — and most companies, holding that list, would rather we kept building it with them.
Reading the matrix
The chart in phase three plots two scores per person — how willing they are, and what they can actually do. Not everyone in the building: the people you put forward as key to this, the ones who need to be able to do it.
The surprises are the point. The person everyone assumed would lead this is often waiting to be convinced. The one already doing it at weekends is usually not in the office. And the owner is very frequently in the overreach box, which is a conversation worth having early.
The cut lines sit at ability 4.5 and willingness 5.5, not through the middle — ability is scored against what someone has actually built with AI, so daily chat use alone cannot climb past 4. Your leadership sees the chart. Nobody sees an individual's answers, and we say so on screen before anyone answers.
Willing and able. Go deep with them early — at month three, these are the people who can take the reins.
All the enthusiasm, not the judgment yet. Left alone they take on more than they should, and it is often the owner.
Entirely capable, not yet interested. Frequently your strongest future advocates, once someone answers the objection they actually have.
Not there yet, or genuinely undecided. We will tell you where they sit and why. What you do about it is a management call, not ours.
How far you take it
These are separable. You are not required to buy all of it, and we will tell you plainly when a layer is not worth it for a company your size.
Assessment, three sessions each, reassessment at the end. This part looks similar in every industry, and it is the layer that pays back fastest.
Policy, configuration, and shaping the platform around how your company actually works. Judgment applied to your specifics — not a template with your logo on it.
The dashboards, tools and workflows that come out of your own idea queue. This is the bespoke layer, and the one we scope with you rather than promise up front.
Once you have a capable team and a tool generating more tools, the hardest questions become yours: what the company is for now, what you sell, who does what. A different conversation, and the one that lasts.
Month three
If the matrix has surfaced someone willing and able, and they want it — the internal owner is the best outcome available. We hand over, stay reachable, and stop billing you monthly.
The queue does not empty; it fills. Most companies keep us on to project-manage and implement what their own people asked for, and to keep training as new staff and new capability arrive. Month to month, so it stays your call every month rather than a year you have to escape.
You keep the policy, the configuration, the training record, the assessment scores and anything we built. Nothing is held hostage, and nothing stops working.
None of this is built to trap you — the handover is real, and so is stopping, which is why it is on the list at all. In practice most companies keep going, because the list of things worth building gets longer rather than shorter, and it moves faster with someone project-managing it who has done it before.