Constantine Consulting

Your first three months, step by step

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

Installing it is the easy part. It's what the tool starts next that needs managing.

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.

We have walked this path with a dozen companies. You have not walked it at all. That is the whole of what you are hiring.

The path

Five phases, and what each one is worth to you.

We understand the business, settle the rules, and design the environment.

Claude's technical settings encode operating choices. We make those choices deliberately — with leadership — before anyone gets a login.

What we do

  • Your principals and owners complete a structured leadership intake covering the business, technology stack, data exposure, access needs and operating priorities
  • We lead an AI-assisted governance workshop using a policy framework to decide approved uses, risk thresholds, human review and how your staff may build with AI
  • Together, the intake and leadership decisions become a Claude deployment blueprint: plan, permissions, connectors, plugins, skills, customizations and training priorities
  • That approved blueprint drives the build in phase two — instead of factory defaults quietly making the decisions for you

Your part

  • Complete the structured leadership intake.
  • Join one working session with the people who set policy and make operating decisions.

Two inputs determine how your Claude environment gets built

Phase I → Phase II
INPUT 01 · BUSINESS REALITY

Structured leadership intake

How the company actually works today.

  • Business and technology stack
  • Data and regulatory exposure
  • Roles and access requirements
  • Goals and operating priorities
INPUT 02 · LEADERSHIP INTENT

AI-assisted governance workshop

How leadership wants AI to operate.

  • Approved and prohibited uses
  • Risk and human-review thresholds
  • Building and automation permissions
  • Long-term operating principles
PHASE I OUTPUT

Your Claude deployment blueprint

The governing decisions and technical specifications we will build from.

AI use policyLeadership-approved guardrails
Plan choiceTeam or Enterprise
Access modelAdmins, roles and seats
Connected toolsConnectors and plugins
Working layerSkills, Projects and customizations
Training prioritiesWhat each role needs first
PHASE II We build the instance before your team logs in.
We do not configure Claude from its defaults. We configure it from how your business actually works and the operating rules your leadership chooses.
View an illustrative AI use policy excerptOne output from the blueprint · fictional company
NORTHVALE CONSTRUCTION · ADOPTED 14 MARCH · V1.0

Artificial Intelligence Use Policy

Applies to every employee, contractor and temporary worker with a Northvale account. Reviewed by the leadership team each quarter.

1 · Approved tools

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.

2 · Never goes into an AI tool

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.

3 · A person reads it before a customer does

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.

4 · Building with AI

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.

5 · Who approves what

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.

What you get

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.

We build the thing before your team ever opens it.

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.

What we do

  • Every setting tuned to the deployment blueprint your leadership approved — retention, memory, what may be shared, which connectors may act without asking first
  • Your company's context written in at the organisation level, so it applies to every conversation anybody has
  • Your standard operating procedures turned into reusable skills your staff can run
  • Your projects and workspaces built out, per team, before anyone is invited in
  • Your owner cockpit stood up on day one — your systems reconciled onto one screen, so the numbers you run the business on stop living in nine places

Your part

  • Create the account and make us an administrator.
  • That is genuinely it.

The owner cockpit, standing up alongside it

Demo · illustrative data
SWIPE SIDEWAYS, OR OPEN IT BELOW → Open the live example cockpit →
A working demo on fictional data. Nine systems — accounting, job costing, CRM, fleet, ad platforms — reconciled onto one screen overnight, with every number labelled by where it came from, and the day's problems ranked by dollars at risk rather than by which system raised them.

What you get

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.

Before we train anyone, we find out who we are training.

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.

What we do

  • Everyone you put forward takes a short readiness assessment: about four minutes, seventeen questions
  • Your coach also sits down with each of them and scores the same two dimensions; the readings are blended, and the human one wins where they disagree
  • You see your key people plotted on one chart, with the four groups named
  • We build the training plan from it with you: who goes deep, who goes in a cohort, who is left where they are for now

Your part

  • Name your key people.
  • Four minutes each from them.
  • One session to look at the chart and make the calls only you can make.

AI Readiness Matrix — Northvale Construction

Illustrative · fictional company
ABILITY CUT · 4.5 WILLINGNESS CUT · 5.5 12 34 56 78 910 12 34 56 78 910 ABILITY → WILLINGNESS → OVERREACH RISK CHAMPIONS ADOPTION RISK SLEEPERS ON THE FENCE Marcus · Ops Manager — willingness 9.2, ability 8.6 (champion) Marcus · Ops Manager Tomás · Field Supervisor — willingness 9.6, ability 7.4 (champion) Tomás · Field Supervisor Naomi · Estimator — willingness 7.4, ability 6.2 (champion) Naomi · Estimator Ruth · Project Coordinator — willingness 6.4, ability 5.4 (champion) Ruth · Project Coordinator Dale · Owner — willingness 9.4, ability 3.2 (overreach risk) Dale · Owner Renee · Sales Lead — willingness 8.2, ability 2.6 (overreach risk) Renee · Sales Lead Colin · Scheduler — willingness 6.8, ability 3.8 (overreach risk) Colin · Scheduler Wes · IT Manager — willingness 4.2, ability 9.2 (sleeper) Wes · IT Manager Aisha · Marketing Lead — willingness 3.0, ability 8.2 (sleeper) Aisha · Marketing Lead Frank · Controller — willingness 2.2, ability 6.6 (sleeper) Frank · Controller Lorna · AP Clerk — willingness 3.4, ability 2.8 (adoption risk) Lorna · AP Clerk Bud · Warehouse Lead — willingness 2.0, ability 3.6 (adoption risk) Bud · Warehouse Lead Hana · Office Manager — willingness 5.0, ability 5.8 (on the fence) Hana · Office Manager Ivan · Foreman — willingness 5.0, ability 3.4 (on the fence) Ivan · Foreman
SWIPE THE CHART SIDEWAYS →
Champions Overreach risk Sleepers Adoption risk & on the fence
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 this chart. Nobody sees an individual's answers, and we say so on the screen before anyone answers.

What you get

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.

Three sessions a person, built around the job they actually do.

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.

What we do

  • Session one: how to think about it. Session two: how it works. Session three: how it works in your role specifically
  • Cohorts where that is efficient, one-to-one where it is not, and extra modules for the people ready to go further
  • In the last few minutes of every session we ask what this has given them an idea for — and we write it down for them
  • Those ideas land in one queue your leadership reviews weekly: approve it, hold it, or turn it into a project
  • We architect what fits together — two departments asking for the same dashboard become one dashboard — and we build what you approve

Your part

  • Roughly three hours a week of your key people's time.
  • An hour a week to set priorities on the queue.
  • Say no sometimes. We will tell you when it is warranted.

The training board — Northvale Construction, week seven

Illustrative · fictional company
AI Training · key people
14 key people 11 baselines taken 3 due WEEK 7 OF 12
MarcusOps Manager W 9 · A 9 1 Onboarding2 Tool use3 Role-based+ Automation
TomásField Supervisor W 10 · A 7 1 Onboarding2 Tool use3 Role-based
NaomiEstimator W 7 · A 6 1 Onboarding2 Tool use3 Role-based
DaleOwner W 9 · A 3 1 Onboarding2 Tool use3 Role-based
AishaMarketing Lead W 3 · A 8 1 Onboarding2 Tool use3 Role-based
WesIT Manager W 4 · A 9 1 Onboarding2 Tool use3 Role-based
+ 8 MORE
The same board your coach and your leadership work from: who has taken the baseline, who is due, and where each person sits in their three sessions. Two scores and a schedule — nobody's answers are on this screen.

What you get

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.

The point of the three months is the decision at the end of them.

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.

What we do

  • We re-run the assessment on everyone we trained, so you see movement rather than a snapshot: where each person started, where they landed
  • We name your champions — the people who came out willing and able, and who want it
  • We hand you the queue in full: what got built, what is worth building next, what should be dropped
  • We tell you plainly whether you still need us, and for what
  • And we are glad to keep going — most clients do, month to month, with us project-managing and implementing the tools their own people asked for

Your part

  • One session to look at the evidence.
  • Decide who owns this internally from here.

The board you and your leadership keep working from

Illustrative · fictional company
ConstantineCONSULTING VIEWING · NORTHVALE CONSTRUCTION dale@northvale.example
AI TrainingAI ProjectsIdea QueueClaude Usage

Projects

Drag to reorder priority.

All stagesAnyone next BoardGantt+ Add project
All time · 3 active — click the timeline to pick a month
PROJECT MAR 26APR 26MAY 26JUN 26TODAY
● NORTHVALE CONSTRUCTION 4 PROJECTS
Materials ordering Building
Weekly executive summary Building
Job cost dashboard Building
COMPLETEBid tracker
Start → MVP MVP → deadline Completed Today
SWIPE SIDEWAYS →
Every approved idea becomes a project with a date, an owner and two phases — a prototype your people can touch, then the live version. The same board runs the work whether your champion is driving it or we are, which is what makes handing it over possible at all.

What you get

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.

PHASE 1 OF 5 · WEEKS 1–2

Reading the matrix

Four groups, and what each one asks of you.

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.

Champions

Willing and able. Go deep with them early — at month three, these are the people who can take the reins.

Overreach risk

All the enthusiasm, not the judgment yet. Left alone they take on more than they should, and it is often the owner.

Sleepers

Entirely capable, not yet interested. Frequently your strongest future advocates, once someone answers the objection they actually have.

Adoption risk & on the fence

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

Four layers. Most companies start with the first two.

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.

1

Training your people

Assessment, three sessions each, reassessment at the end. This part looks similar in every industry, and it is the layer that pays back fastest.

2

Implementation

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.

3

Building what you need

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.

4

Executive coaching

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

Three ways forward, and you pick with evidence.

1

Hand the reins to your champion

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.

2

Continue month to month — what most clients do

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.

3

Stop here

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.

johnconstantine.ai