Rebuild the operating model behind the customer experience, not just the interface.
Customer experience degrades where operations, not screens, decide what the customer receives.
Boards are redirecting support and marketing budgets into automation faster than anybody is measuring the return, and the pilot that reaches a customer is the one nobody gated.
In your organisation, that shows up as a platform proposal you are holding, or a pilot you have extended twice. The tooling is real-time, and your people still decide weekly, so the speed you paid for never reaches the customer.
Three modules answer the three questions underneath that.
Those questions are how your organisation decides, what personalisation is worth in your revenue, and whether a deployment already running should continue. Which one you take depends on the question that is live for you.
Across all three, I design how your organisation decides, I build the value model, and I hold the gate. I do not build the system.
Real-time tooling, weekly decisions
What I build and what I do not
Your engineers build the system, or your systems integrator does, or your platform vendor does. That limit is permanent, because it is the honest shape of what one person can carry.
The separation matters commercially as well. Building a decisioning platform runs into six figures and two quarters.
The person who designs the gate should not also be the person whose invoice depends on getting through it. That separation is worth something to you independently of who I am.
If you want a single supplier to handle design and build, buy that from somebody with a delivery bench.
Where the money goes wrong
Failure
What it costs
Platform bought before the value model exists
Somebody writes the business case backwards to justify a committed spend
Campaign teams handed real time tooling
Your people still decide weekly, so the speed you paid for never reaches the customer
Personalisation measured on engagement
The team claims lift on a measure that moves while revenue stays flat
Nothing written down that would end the pilot
The pilot runs on forever, because nobody defined what would end it
Model marking its own homework
The system that generates the recommendation also scores whether it was good
The last row is the one a vendor is least likely to raise with you.
Nothing marks its own homework.
Every system here now has a model inside it, so the question worth asking is where that model sits relative to what judges its output.
When the model that generates a recommendation also scores whether that recommendation was good, you have a machine marking its own homework. These models flatter work put in front of them, and their own output hardest of all, so a second model in the judge’s seat does not clear it.
The only judge worth having is one nobody can talk round. That means a check giving the same answer every time, run by something other than the system being scored.
C3 is where I write those checks, and it carries how they hold. I hold the position because I built it into my own product and watched it hold.
I was CX Solution Lead and ran the Generative AI practice at EY UK and Ireland, and built it into a three million pound book of business. The four-dimensional gate I designed for a UK national broadcaster’s Generative AI Discovery holds risk inside the value score so that nobody can trade it away separately.
They still run it.
Which module you need
Module
For
Who buys it
Duration
C1
A platform acting in a second against a weekly decision cycle
A CMO, a chief customer officer, or a head of customer operations
6 to 10 weeks
C2
A platform proposal whose arithmetic the vendor wrote
A chief revenue officer or a CFO holding the proposal
4 to 6 weeks
C3
A pilot nobody gave a condition it could fail
A COO or a transformation lead facing a third extension request
3 to 5 weeks
C1 designs how your organisation decides at the moment of contact. C2 models what personalisation is worth in your own revenue, and C3 writes the gate the deployment should have carried from the outset.
Each page carries its own phases, its own limits and what it settles.
I usually run C2 first, even though it is numbered second, because a value model built after a platform commitment is a justification exercise.
You can also take activities inside each module individually. I set the fee after the first conversation. It scales with the number of segments and contact channels in scope, and with how much baseline data your team already holds.
What all three modules produce
You get
What it changes
A figure worked out in your own revenue
The platform decision stops resting on the vendor’s model
Decision rights at the point of contact
Your people decide in the second the customer is on the line
A check run by something other than the system
The recommendation stops being scored by whatever produced it
Written conditions that halt a phase
Continuing becomes a decision on a date, with a name against it
The assumption the case dies on, named
Your board argues about the assumption while it is still cheap
What the evidence covers
Customer operations is the record. At a UK enterprise telco, I turned CX craft into a 204-page operating manual. I embedded three mandatory checkpoints into the existing project governance, with KPIs tied to customer commitments.
At a global standards body, I blueprinted the service end to end, and at a UK energy retailer I moved self-service to a search-first channel. The value modelling behind C2 sits on my record in adjacent form.
The gate is work I have built and shipped in my own product, and I have built and tested the checking work behind C3.
So judge these three on that design work, and on how hard I push its assumptions. Deployment into live customer operations is your own delivery teams’ work.
I quote no figures from the classifier study behind the gate, because its predictors were partly inputs the gate already used. C3 carries that reasoning in full.
Where to start
If your problem is a platform decision with no value model behind it, then Personalisation Business Case is for you, and it is the place I would start.
If the value model holds and you have no function to govern the programme, then AI Operating Model Design is for you. Where licence and seat economics are the live question, AI Value Realisation is for you.
C1 - Customer Engagement Operating Model Design moves the engagement decision to the moment of contact and names the unit that owns the outcome. Your platform can act in seconds while the people who decide how it acts meet on Tuesday mornings, so an agent who can see a recommendation is wrong for the customer either reads it out or quietly ignores it, and ignoring it costs them less. With the Customer Engagement Operating Model Design module, we settle who decides what is allowed inside the window a call allows. We also map the route your people follow when the automation is wrong.
C3 - Deployment Stage Gate Review is the gate your deployment should have carried from the outset, written before the next stage opens. The extension request has reached you for the third time, and everyone on it can produce genuine evidence that your deployment is working, because the team that built the system also designed the dashboard and chose the measures after the behaviour was visible. With the Deployment Stage Gate Review module, we set the conditions for continuing and for stopping, and the checks that test them run outside the system they judge.
C2 - Personalisation Business Case is the value model for personalisation, built on your own revenue and finished before you commit to a platform. A proposal has landed where every figure is defensible in isolation, and none of the benchmarks describes your customers, your margins or your retention curve, and the party that benefits from the signature wrote the arithmetic. With the Personalisation Business Case module, we price the lift available per segment, the churn a better-timed contact could save, and the economics per resolved contact.
