What I do in the first quarter. Before anything is built.

1. Days 1 to 14

   Measure what nobody is measuring

   Two weeks is enough to find the truth, and it is not in the dashboard.

   • Listen to and read raw contact, unfiltered and unsampled, across every channel the operation runs. Not a sample deck prepared for me. The actual conversations, including the ones that went badly.
   • Count the contacts that should never have happened: the chase, the repeat, the transfer, the status check, the apology for a failure elsewhere in the business. That number is almost always the largest single line in the cost base and almost never on the report.
   • Walk the floor and sit with the advisors who are good at this, and ask each of them the same question: what do you have to work around every day.
   • Establish the baseline the board will be shown against later: answer times by channel, resolution, repeat contact, conversion where a contact carries revenue, attrition, and what a customer who complains is worth twelve months on.

   By the end of it A baseline signed by finance, and a ranked list of the contact the business is paying for and should not exist.

2. Weeks 3 to 6

   Decide, and say no in public

   By week six the operation should know what is changing and, just as importantly, what is not.

   • Pick the two or three moments where the money actually moves, from the evidence rather than from the roadmap, and stop work that is not aimed at them.
   • Write the data contract: what has to arrive in the conversation, live, for each of those moments, and which system owner owns each field. Most operations already own the platforms. What is missing is a document precise enough that each owner knows exactly what to build.
   • Agree the policy gates with legal, risk and brand before anything is built, not after: what may be answered automatically, what must be disclosed, what must reach a named human, and what is refused.
   • Name the people. Every moment that matters gets an owner with a name, not a function, and they are told what they are accountable for.

   By the end of it A signed data contract, agreed policy gates, and a public list of what this operation has stopped doing.

3. Weeks 7 to 12

   Live, measured, and earning

   Something real is in front of customers by week twelve, on the technology already owned.

   • Build on the vendors already paid for. No new licence until something the business can measure has earned it.
   • Start in shadow: real contacts read, real answers drafted, humans approving every one, and the difference against the baseline measured daily rather than quarterly.
   • Go live on the narrowest honest slice, one product line or one channel set or one team, alongside everything that exists today and never instead of it.
   • Read it weekly against the week two baseline, in the open, including the weeks it does not move.

   By the end of it Live customer-facing capability, the operation's own numbers rather than modelled ones, and a board-ready read on whether to scale, adjust or stop.

The gates are held honestly. If the baseline cannot be established by week two, that is the finding and the board hears it. If the shadow read is not beating the baseline by week nine, it is fixed or it stops. At week twelve you scale only what the numbers earned, and you keep the baseline, the data contract and everything built either way.

• I will not open with a platform selection. If the answer to a customer operations problem is a procurement exercise, the problem has not been diagnosed yet.
• I will not promise a percentage in month one. The baseline in week two is what makes any later number honest, and before it exists any figure I give you is decoration.
• I will not put a model in front of customers without a named human who owns what it says. Automation without accountability is how operations lose the trust they spent years earning.
• I will not run a proof of concept that cannot survive contact with a real customer. Sandboxes prove the vendor works, not that the operation does.
• I will not hide a bad week. If the shadow read is not beating the baseline by week nine, the board hears it in week nine.