The NATIVE framework
Six dimensions decide whether agentic AI holds inside a company. We score them across five maturity levels before we talk about tools, and we point the roadmap at the weakest one.
- North Star
- Architecture
- Talent & Teams
- Integrity
- Value Tracking
- Execution & Scale
A maturity model that measures what your company needs in place before AI agents can take over real work
The NATIVE Assessment is the engagement where we do the measuring. Four to eight weeks, fixed price, ending in a report with your score, your maturity level, the area holding you back and a costed plan for fixing it. The framework behind it carries no price and stays in use through every project that follows.
An assistant drafts something for a person to check and send. An agent does the task itself: it reads the enquiry, looks up the customer, applies your pricing rules, sends the quote and records what it did.
That is why maturity matters. An assistant needs little more than a licence, while an agent needs write access to your systems, rules somebody defined, a person accountable when it goes wrong, and a record of every action it took. NATIVE measures whether those conditions exist in your company.
N — North Star
A business goal with a number, a date and a name against it: which process, how much faster or cheaper, owned by whom. At L0 the goal is "do something with AI"; at L4 every agent traces back to a target somebody is accountable for.
A — Architecture
The state of the data and systems underneath: where each data object lives, which system leads it, whether interfaces are documented, and whether an agent could read and write without a human copying between two tools. This is the dimension that comes out lowest in most assessments.
T — Talent & Teams
Who works with the system and who is trained for it: which roles change, who reviews an agent’s output, who is allowed to intervene. Scoring high here does not mean everyone sat through a training — it means the new tasks are written into job descriptions.
I — Integrity
Traceability, approvals and data protection: whether you can reconstruct why an agent did something, who signed it off, and which data it touched. In a regulated process this has to be in place before go-live, not after.
V — Value Tracking
Whether the value case is still measured after go-live: the baseline before, the number after, and who reports it. Without the baseline there is no way to show afterwards what the rollout achieved — the most common reason one cannot be defended.
E — Execution & Scale
Operations and expansion: who owns the running system, how a failure surfaces, and what it takes to add the next use case. At L4 this is routine rather than a project.
Built for companies that have already tried AI and now want it running a process.
NATIVE is for companies at the point where tools are in daily use but nothing completes a process end to end, where a budget decision is coming and the business case has to hold up, where the same process runs in several countries and needs one comparable picture across them, or where the process is regulated and governance has to be settled before go-live.
Five levels, and what each one means for the business.
A company that is far along in five of six and at zero in one is at zero — which is why the roadmap starts there rather than where the work is most fun.
L0
Manual
The work runs by hand, AI does not appear in it.
L1
AI-Assisted
Individuals use tools, with no rule behind it and no effect on the process.
L2
AI-Augmented
AI sits inside the process, but the data foundation does not carry it.
L3
Agentic
Agents run, but operations, adoption and measurement are missing.
L4
Agent-Native
Agents carry whole role scopes, across units and countries.
A score, a level, and the one dimension to fix first.
The assessment returns both figures together. A company can score well overall and still have no agent in production, because the weakest dimension is the one that sets the level, and it is where the roadmap starts.
NATIVE Score across six dimensions
Maturity level, set by the weakest
You come out with a score, a bottleneck and a plan.
We collect evidence from your documents and systems, interview the people who actually run the process, and score every dimension. A separate data module looks at where your data sits today and what it would take for an agent to act on it. Out of it come your score and level across all six dimensions, a gap analysis of the weakest one, a target data architecture, a business case built on a baseline recorded before anything changes, and a roadmap that starts at the bottleneck.
Four to eight weeks
Kick-off to final readout, scoped at the start.
Evidence based
Documents, systems and interviews. Not a questionnaire.
