---
title: "IMPACT AI™: the methodology for adopting AI with evidence, phase by phase"
description: "How to adopt AI without stalling at the demo: six phases with decision gates, the AI Readiness Index™ and data maturity per UNE, quick wins, pilots with KPIs, governed adoption and shadow-based handover."
url: https://bidatia.com/en/ideas/metodologia-impact-ai-adopcion-ia
lang: en
date: 2026-10-27
updated: 2026-11-04
author: "José Luis García"
keywords: ["ai adoption methodology", "impact ai", "data maturity une 0080", "dama-dmbok", "ai quick wins", "ai mvp", "shadow reverse shadow", "ai strategy"]
publisher: Bidatia
---

# IMPACT AI™: the methodology for adopting AI with evidence, phase by phase

> How to adopt AI without stalling at the demo: six phases with decision gates, the AI Readiness Index™ and data maturity per UNE, quick wins, pilots with KPIs, governed adoption and shadow-based handover.

## Executive summary

- IMPACT AI™ organises AI adoption into six phases (Identify, Measure, Prioritize, Activate, Consolidate and Transfer), each closed with a gate: a control question, minimum evidence and a documented decision.
- Readiness is measured with the AI Readiness Index™ across six dimensions, and data readiness with UNE 0077, 0078 and 0079, assessed with UNE 0080, and the DAMA-DMBOK framework.
- Opportunities are scored on value, feasibility, adoption, effort and risk; quick wins approached as MVPs are prioritised, with a GO, conditional GO, PARK or NO GO decision.
- Each pilot has a Pilot Charter, training applied to the use case and its own control dashboard; adoption is ensured through human oversight, AI Governance Lite™ and a Success Dashboard.
- Handover is done with shadow and reverse shadow and an AI Action Plan™: the work ends when the client's team is in control.

Many organisations have already tried AI: an assistant in one department, a proof of concept that impressed in the demo, Copilot licences handed out without a plan. And yet few can say how much value it has delivered. The reason is hardly ever the technology. It is that they started with the tool rather than the business.

**IMPACT AI™** is the methodology we use at Bidatia to support organisations from the first conversation about AI until they use it in their processes, measure its results and govern it without depending on us. In this article we explain how it works, phase by phase. Full details are on the [methodology page](/en/methodology).

## Why so many AI projects never get past the demo

The symptoms are always the same:

- **The tool is chosen first** and only then does anyone look for a use for it.
- **Nobody has measured the starting point**, so it is impossible to prove any savings.
- **The data is not ready**: duplicated, out of date or without an owner.
- **The pilot has no success criteria** defined before it starts, and it never quite gets closed.
- **The supplier leaves and the knowledge leaves with them.**

IMPACT AI™ is designed to avoid each of these pitfalls. Its core rule is simple: we do not start with which tool to use, but with the business, the problem, the process, the people, the data and the expected outcome. The technology comes later.

## Six phases, six decision gates

The name sums up the six phases: **I**dentify, **M**easure, **P**rioritize, **A**ctivate, **C**onsolidate and **T**ransfer. Each phase ends with a **gate**: a control question, minimum evidence and a documented decision. This is the difference between activity and progress: a meeting, a workshop or a prototype are activities; passing a gate requires evidence.

| Gate | Control question | Minimum evidence | Decision |
|---|---|---|---|
| G1 Identify | Do we understand the problem and the opportunities? | AI Opportunity Canvas™ and pain points | Move on to measuring |
| G2 Measure | Is the company ready? | AI Readiness Index™ and gaps | Correct / continue |
| G3 Prioritize | Which use case deserves to go first? | Portfolio and assessment of use cases | GO / NO GO |
| G4 Activate | Does the solution prove its value? | Pilot results against the KPIs | Adopt / iterate / stop |
| G5 Consolidate | Is it being used and delivering results? | Usage, efficiency, business and adoption | Consolidate |
| G6 Transfer | Can the company carry on by itself? | Backlog, owners and roadmap | Scale / operate |

## Identify: understanding the business as the people doing the work experience it

We talk to senior management, but above all to the people who do the work every day. That is where the gap shows between what should happen and what actually happens: repetitive tasks, waiting times, rework, decisions taken late or with little information, data copied by hand from one system to another. Each problem becomes a testable hypothesis: *"if we apply this capability to this process or this data, we could improve this outcome for this user"*. Still without choosing a tool or a supplier. In this phase we also take stock of the AI already in use across the organisation, including the AI built into applications, which tends to be more than people think. We explain this in [the AI you already have without knowing it](/en/ideas/ia-integrada-en-aplicaciones-de-terceros).

## Measure: assessing readiness with evidence, not an opinion

An impression is not enough to know whether an organisation is ready. The starting point is set by the **AI Readiness Index™**, which assesses six dimensions: **strategy** (objectives, sponsor, decision-making capacity), **processes** (digitalisation, stability, indicators), **data** (availability, quality, access, ownership), **technology** (applications, integration, cloud, AI tools), **people** (skills, current usage, capacity for change) and **governance and risk** (privacy, security, control, oversight). Each score is backed by evidence; anything that cannot be checked is recorded as "not verified".

The result places the organisation at one of five readiness levels, each with its practical implication: **Explorer** (start small and train), **Starter** (quick wins with support), **Ready** (activate the priority use cases), **Adopter** (consolidate and expand) and **Scalable** (a portfolio of use cases and scaling up). More than the figure itself, what matters is explaining what evidence justifies each level.

We measure the data dimension against a standard: the **UNE 0077** (data governance), **UNE 0078** (data management) and **UNE 0079** (data quality) specifications, assessed with **UNE 0080**, and the **DAMA-DMBOK** framework.

UNE 0080 assigns each process a maturity level:

| Level | What it means | What it takes to move up |
|---|---|---|
| 1. Performed | Data processes are carried out, but reactively and dependent on specific individuals | Identify critical data, appoint owners and apply the first rules |
| 2. Managed | They are planned and tracked in specific areas or projects, with identified owners | A corporate model: common policies, roles and processes |
| 3. Established | They are standardised and documented across the whole organisation | Measure them with indicators, targets and thresholds, and analyse the causes of deviations |
| 4. Predictable | They are managed with quantitative indicators and their outcome is predictable | Embed continuous improvement and innovation |
| 5. Innovating | Data is managed to obtain maximum value and adapts to new objectives | Sustain the level and extend it to new domains and AI initiatives |

The result is not a score, but a prioritised list of gaps: what needs to be closed before each AI use case. If data is holding things back, we include it in the roadmap and address it in parallel with the first use case. To find out more, see [data quality: dimensions and metrics](/en/ideas/calidad-del-dato-dimensiones-metricas) and [data governance: roles and organisation](/en/ideas/gobierno-del-dato-roles-y-organizacion).

## Prioritize: choosing the right place to start

We score each opportunity from 1 to 5 on five factors: **value** (savings, revenue, quality or service), **feasibility** (data, tools, integration and skills), **adoption** (sponsor, users and willingness to change), **effort** and **risk** (privacy, security, legal, reputation, reliability). The score does not replace professional judgement: it forces a structured conversation and puts the reasons in writing. For the shortlisted use cases we prepare a light business case, in which the expected improvement is a hypothesis, not a promise, and we distinguish which data has been measured and which is estimated. And we prioritise two things:

- **Short quick wins**: use cases lasting a few weeks that deliver visible value to the business. They help show results early and build the organisation's confidence.
- **MVP approach**: each use case is scoped to the minimum needed to validate the hypothesis with real data before investing further.

The sponsor decides on three to five shortlisted use cases, and each one ends with a documented decision: **GO**, **conditional GO** (there is value, but an enabling action is needed first), **PARK** (interesting, but not the right time) or **NO GO**. Dropping a use case that does not pay off in good time is also a good outcome.

## Activate: piloting with control, not faith

The prioritised use case is designed in writing (problem, user, current flow, solution, data, scope, KPIs, risks and owner) and formalised in a **Pilot Charter** before starting: hypothesis, users, duration, metrics, success criteria and stop condition. Training is tied to the use case: we teach the essential concepts, the limits of AI and human review using tasks from the process itself, and it is validated with a real task, not by attendance.

The pilot goes into production with a controlled scope and is governed from day one. Steering committees are necessary, but not sufficient. That is why each pilot has its own control dashboard:

- **Steering committees**: an executive one for decisions, priorities and risks, and an operational one for progress and issues.
- **Quick win control**: a baseline measured before starting and business KPIs (hours, errors, cost per transaction, sales) with agreed success criteria.
- **Project control**: time, cost and scope against plan, quality of answers or predictions, and actual user adoption.

A use case does not "work" because a demo is impressive. It works if there is evidence against the defined KPIs, and that evidence drives the decision: adopt, iterate or stop. We describe the full life cycle of an AI project, with its controls, in [the life cycle of an AI project](/en/ideas/ciclo-de-vida-proyecto-ia).

## Consolidate: making sure it is used and keeps working

A solution that works but nobody uses adds nothing. In this phase we build the change into the way people actually work:

- **Adoption Plan**: affected users and champions, which step of the process changes and what exceptions there are, training, support, communication and follow-up.
- **Human oversight**: what the AI produces, what a person reviews, which decisions can be automated and which thresholds require escalation.
- **AI Governance Lite™**: purpose, owner, users, data, risks, controls and review. Proportionate governance, without unnecessary red tape.
- **Success Dashboard**: usage, efficiency, business, adoption and risk indicators.

If the use case is not being used, we first look for the barrier; only then do we talk about scaling.

## Transfer: our work ends when you are in control

Handover is not a final meeting with a document. We apply **shadow** and **reverse shadow**:

1. **We operate**, with your team kept informed from the outset.
2. **Shadow**: your team works alongside us and sees how we operate, make decisions and resolve issues.
3. **Reverse shadow**: your team operates and we supervise, correct and answer questions.
4. **Full control**: your team operates on its own, with documentation, a prioritised backlog, owners and a roadmap.

Each profile comes away with what it needs to be able to do: senior management understands results, risks and the next investment; the use case owner runs the KPIs and decides on improvements; users handle the exceptions; and the technical profile manages configuration, access and integrations. We leave an **AI Action Plan™** with the prioritised backlog, the outstanding data capabilities and training, the controls to be maintained and the next milestones, and we repeat the readiness assessment to compare it with the starting point.

The project closes when your team operates autonomously, not when the timeline runs out.

## Learning by doing

Training is not a course at the start: it runs through every phase. In Identify, people learn to spot problems suited to AI; in Prioritize, to think in terms of value, feasibility and risk; in Activate, to use the chosen capability on real tasks; in Consolidate, to operate it responsibly; and in Transfer, to carry on without a consultant. The final test is simple: each participant must be able to explain the use case, carry it out, recognise when not to trust the result and know who to escalate an exception to.

## The six principles we do not compromise on

- **Business before technology**: we do not propose tools before understanding the objective and the process.
- **Evidence before opinion**: we separate observed facts, perceptions and hypotheses.
- **Value and adoption**: a technically feasible use case that nobody will use is not a priority.
- **Proportionate risk**: controls matched to the risk and size of the organisation, without red tape.
- **Experiment before scaling**: we test hypotheses with a limited scope and KPIs.
- **Transfer**: every engagement increases in-house capability, without creating dependency.

## Do you have to go through every phase?

Not necessarily. We join at whatever point each organisation has reached. If you already have your use cases prioritised, for example, we start at Activate, first making sure the controls from the earlier phases are in place: success criteria, available data and assessed risks. And regulation is not left until the end: the requirements of the EU AI Act and the GDPR are taken into account from prioritisation onwards, as we explain in [which regulations apply to AI and data](/en/ideas/ia-y-datos-normativa-rgpd-ria-ens).

If you would like to know which phase you are in and where we would start, [tell us about your case](/en/#contact).

## Frequently asked questions

### What is IMPACT AI™?
It is Bidatia's methodology for adopting AI in six phases (Identify, Measure, Prioritize, Activate, Consolidate and Transfer), each closed with a gate: a control question, minimum evidence and a documented decision.

### How do you measure whether a company is ready for AI?
With the AI Readiness Index™, which uses evidence to assess six dimensions: strategy, processes, data, technology, people, and governance and risk. The data dimension is measured with the UNE 0077, 0078 and 0079 specifications, assessed according to UNE 0080, and the DAMA-DMBOK framework.

### What is a quick win?
A short use case, approached as an MVP, that delivers visible and measurable value to the business and validates the hypothesis before investing further.

### How do you know whether an AI pilot is successful?
With a Pilot Charter and success criteria set before starting, a measured baseline, business KPIs for the quick win and project control over time, cost, quality and adoption, reviewed in the steering committees.

### What are shadow and reverse shadow?
Two stages of the handover: in shadow, the client's team accompanies Bidatia's team while it operates; in reverse shadow, the client's team operates and Bidatia supervises, until the client is in control.

### What happens if a use case is not worth it?
The decision is based on evidence: at prioritisation it can be marked as conditional GO, PARK or NO GO, and after the pilot the decision is to adopt, iterate or stop. Dropping a use case in good time is also a good outcome.