AI Agents
All insights

Playbook

How we sign a KPI Contract

If the pilot does not move the KPI we agreed on, there is no scale-up and no retainer. The four parts of the clause, an anonymized example and the things we refuse to sign.

Mauricio Moreno

Co-Founder & Co-CEO

5 min read
A one-page agreement with one figure circled in violet ink on a walnut desk

A KPI Contract is a clause we sign before building: if the pilot does not move the KPI we agreed on, there is no Scale phase and no retainer. It turns the pilot into a measured experiment with a decision at the end, the commitment gate, instead of a project that quietly becomes permanent.

Why we put the number in the contract

Gartner predicts that more than 40% of agentic AI projects will be cancelled by the end of 2027 (Jun 2025) and estimates that only about a quarter of AI use cases in customer service produce ROI (Aug 2026). Most of those projects were not cancelled because the technology failed. They were cancelled because nobody could prove it worked. A KPI Contract makes proof the default.

It also protects us. A pilot without a number is a pilot that can be judged on feelings. We prefer to be judged on a baseline, a target and a date.

The four parts of the clause

  • The KPI: one primary metric, sometimes one guardrail metric. Autonomous resolution rate, cost per ticket, days sales outstanding, completed calls per hour. Never more than two.
  • The baseline: measured by us together with the client during Blueprint, over a window long enough to be honest, typically the previous 8 to 12 weeks of the client's own data.
  • The target: a number and a date. Not “improve”. A specific figure at the end of the pilot, with the measurement method written down next to it.
  • The consequence: if the target is not met at the commitment gate, the client does not pay for Scale and does not sign the retainer. The pilot fee stays paid; the rest does not happen.

An example, anonymized

From a fintech pilot in México. Primary KPI: cost per resolved tier 1 ticket. Baseline: $6.40 USD, measured from the client's own ticketing data before the pilot. Guardrail: customer satisfaction at or above baseline. Measurement: the client's ticketing system and the Agent Console, reviewed weekly. Outcome at the week 12 gate: $0.85 per ticket and NPS from 42 to 71 (Proof Ledger, Sep 2026). The gate was passed. The retainer was signed.

$6.40 to $0.85

cost per resolved tier 1 ticket, USD

Proof Ledger, Sep 2026

42 to 71

NPS over the same twelve weeks

Proof Ledger, Sep 2026

How we get to a number we can sign

The KPI Contract is signed at the end of Blueprint, the paid discovery phase of 1 to 2 weeks. That phase exists so that we can sign responsibly.

  • Workflow map: where the process starts, where it ends, who touches it and where it leaks.
  • Data assessment: does the system of record have what the agent needs, and can we write back to it?
  • Architecture and ROI model: what it costs to run, what it saves, at what volume, on which model.
  • Evals plan: how we will know, every week, whether the agent is getting better or worse.

If Blueprint shows that the data cannot support the KPI, we say so and propose a different process. 85% of the companies that pay for discovery go on to build with us (DaCodes commercial plan, Apr 2026; we are turning that into a measured metric). The other 15% keep a report they can use with anyone.

What we will not sign

  • A KPI the client cannot measure today. If there is no baseline, we build the measurement first.
  • A target that depends on a system we are not allowed to integrate with.
  • More than two metrics. Three KPIs is a dashboard, not a contract.
  • A pilot shorter than 8 weeks. Agents need real traffic and at least one policy change before anyone should trust them.
  • A number picked to look good in a press release. The target has to be one the operations team believes.

What happens at the commitment gate

Week 12, one meeting, one page. The baseline, the target, the measured number, the eval trend, the cost to run. Three outcomes are possible: scale, extend the pilot for a defined reason and a defined period, or stop. We have stopped pilots. It is cheaper for everyone than pretending.

What it changes in practice

The contract changes behavior on both sides from week one. Our pod stops optimizing for the demo and starts optimizing for the number. The client's team stops measuring the agent on anecdotes, because the measurement is agreed and automated. And the executive sponsor gets a single page at the end instead of a sixty-slide deck. That is the whole point.

We sign the number before we start. It is why our discovery is the most honest conversation you will have about AI this year.
kpi-contractpricingpilotcommitment-gate

Mauricio Moreno

Co-Founder & Co-CEO

5 min read

Insights

Related articles

Hands sorting folders of physical records in an archive room
Playbook

Data first: why we assess your data before we build an agent

An agent is only as good as the system of record it reads and writes. What the data assessment in Blueprint checks, what we typically find, and what to bring to the first session.

Alejandro Hynds

Integrations and data

5 min read
A woman checking her banking app at a crosswalk on Reforma in Mexico City

$6.40$0.85

cost per resolved tier 1 ticket, USD · Proof Ledger, Sep 2026

Case

Case: a fintech in México cut cost per ticket from $6.40 to $0.85 in 12 weeks

Tier 1 support moved to an enterprise agent on WhatsApp and web chat connected to the core. Cost per resolved ticket from $6.40 to $0.85 USD and NPS from 42 to 71. Client anonymized; figures verified in the Proof Ledger.

Sarah Mousa

Content and Proof Ledger

5 min read

Which process do we automate first?

Two discovery sessions, no cost, and you leave with two or three quick wins whether or not you build with us.

We reply in under one business day.

Loading the form