The Support Maturity Model.

The Support Maturity Model is a framework for assessing a customer support operation. It places the operation on five stages, from reactive at stage one to strategic at stage five, and scores six parts of it, called lenses, one at a time. AI enters at stage four, once the team has written workflows for it to sit in, maintained content for it to draft from and a quality score to check it against.

By Colin Moroney, co-founder of Manic Edge · Published 15 October 2026

What the model measures

Support teams get asked the same question in different forms: are we ready for AI, why is quality slipping, why does headcount keep rising. Each version is a question about how well the operation is built, and the model answers it in two parts.

The stages describe how the operation runs, from work held together by a few experienced people up to a team whose data shapes decisions outside the department. The lenses split the operation into six parts that can be assessed separately, because teams rarely sit at one stage across the board. A team can have an integrated helpdesk at stage three and a knowledge base that lives in people’s heads at stage one. The overall stage is the average of the six lenses, and the gaps between lenses usually tell a team more than the average does.

The five stages

01Reactive

Every support team starts here. The work gets done because a few people know how, and that knowledge lives with them. Cases are handled differently depending on who picks them up, and quality is judged by feel.

02Structured

The basics are written down: who does what, how a case is meant to move, and where the answers live. A helpdesk is configured and macros are in use. What the team does not have yet is a way to tell whether any of it is working.

03Measured

The operation can be measured. The team knows where the queue stalls, what good looks like, and where the knowledge base falls short. This is the base AI gets added to, and most teams being asked to add AI have not reached it yet.

04Augmented

AI does part of the work: routing, drafting, deflecting or scoring. The question at this stage is proof. The team should be able to show what the AI did last week, what it got wrong and who checked it.

05Strategic

Support feeds decisions outside the department, including product, roadmap and staffing. The knowledge base answers customers directly, and the quality score flags problems before customers report them. Few teams sit here.

The six lenses

Each lens comes with one question that tends to place a team quickly. The answer usually shows up in the artefacts the team uses every day: the escalation list, the macros, the QA form, the dashboard nobody opens.

People & Roles

Who owns escalations, and who decides staffing?

At stage one nobody owns a type of case, so whoever is free takes it, and escalations go to the most experienced person by habit. By stage three, roles and tiers are written down and headcount is set from volume and skill mix, which makes the routine work visible. That routine work is what AI takes on at stage four.

Process & Workflow

Do two agents handle the same case the same way?

At stage one, how a case is handled depends on who picks it up, and there is no written version to compare against. By stage three, workflows are written down, followed on every shift and tracked, so the team can see where cases stall. That defined path is what lets AI take on routing and triage later.

Knowledge & Content

Where does an agent look first, and do they trust it?

At stage one, answers live in people’s heads and leave when those people do. By stage three there is one documented source, and the team tracks the questions it fails to answer and closes those gaps first. This is the lens AI relies on most, because AI drafts from this content just as agents do.

Metrics & Quality

How many interactions were QA scored last month?

At stage one, quality is judged by feel and the numbers count volume, so any change to the team gets judged on volume. By stage three, quality is scored against a fixed rubric applied the same way every time, and the trend gets acted on. That score is what an AI deployment is later checked against.

Tooling & Automation

Which tools are paid for and which are used daily?

At stage one the team works from a shared inbox or a basic ticketing tool and cannot see the queue as a whole. Stage two, a configured helpdesk with macros, is usually the cheapest climb in the model. By stage three the stack is integrated and the reporting can be trusted, so deflection and agent assist can be added and measured.

Customer Insight

When did support data last change a product decision?

At stage one there is no feedback loop, so support hears about a product problem weeks before the roadmap does and the lead goes unused. By stage three, survey results and contact reasons are grouped into themes and reported upward on a schedule. At stage five, support insight sets part of the product roadmap.

The model on one page

Each cell describes one lens at one stage. Read down a column to follow one part of the operation as it develops, or across a row to see everything a stage asks for. The highlighted row is stage four, where AI enters.

Why AI enters at stage four

AI tools depend on three things a support team builds in the first three stages. They need a written workflow to sit inside, maintained content to draft from, and a quality score their output can be checked against. A team at stage three across all six lenses has all of them.

Where a team sits decides the next move. Below stage three, the work before AI is getting every lens to stage three. At stage three across the board, the usual first step is agent assist in the knowledge lens, because a person sees the output before a customer does. At stage four, the question becomes audit: for each lens, can the team show what the AI did last week, what it got wrong and who checked it?

For how to check what an AI agent actually resolves, see deflection rate versus resolution rate.

Which lens to climb first

One climb is one lens moving up one stage. A team at stage one on every lens is twelve climbs from a stage-three base and twenty-four from the top of the model. An assessment counts the climbs and, more usefully, puts them in order.

The first lens to move is the lowest one. When several lenses tie at the lowest stage, the model climbs them in this order:

1. Process & Workflow, because every metric and every AI tool needs a written workflow to sit inside.
2. Knowledge & Content, because AI drafts from this content, and so do your agents.
3. Metrics & Quality, because nothing else can be checked until quality has a number.
4. People & Roles, because someone has to own each of the other climbs.
5. Tooling & Automation, because tools bought before the process is written down get configured to whoever argued loudest.
6. Customer Insight, because it depends on the other five being in place.

Where teams get stuck

The most common stall is stage two to three on Metrics & Quality. The team has a QA form and reviews run every week, but two reviewers can score the same interaction ten points apart. Nobody tracks the trend, so nobody acts on it. Inside the team, QA feels like grading: agents dispute their scores, leads stop trusting the numbers, and the dashboard goes unopened. What gets them moving is calibration. Reviewers score the same five interactions separately, compare results and tighten the rubric until they agree. Then the team picks one quality trend a month and changes something because of it.

The second is stage two to three on Knowledge & Content. A knowledge base exists, but nobody owns it, so articles go stale and agents stop looking there first. From the inside it looks like a busy Slack channel where the same questions get asked every week, and the answers go to whoever replies fastest. Things start moving when someone owns the content and the team logs every question the knowledge base couldn’t answer. The most frequent gaps get closed first.

The third shows up earlier, at stage one to two on People & Roles. Escalations go to the most experienced person out of habit, and that person becomes the system. The team runs fine until they take a week off. Writing the escalation path down, who owns which type of case and when it moves up, usually takes an afternoon, and it’s the climb teams put off longest.

What skipping a stage costs

A support team of about 25 agents was asked to cut contact volume, and the answer it chose was a customer-facing AI assistant. On paper the team looked ready: a configured helpdesk, macros in use, a knowledge base of over 100 articles. Measured against the model, Knowledge & Content was at stage two and Metrics & Quality sat between one and two. Nobody owned the articles, and quality wasn’t scored against anything fixed.

The deflection numbers looked good for the first six weeks or so. Then repeat contacts started climbing, and agents found themselves correcting answers the assistant had given customers a few days before. Nobody could say how often it was wrong, because there was no quality score to check its answers against. The assistant was drafting from content that hadn’t been maintained, so it repeated the old errors faster and to more people.

Recovery meant stepping back a stage. The team switched the customer assistant to an agent assist, so a human agent saw every answer before the customer did, audited the top 50 articles by contact volume and assigned owners, and built a fixed QA rubric. The assistant went back to customers about three months later, and this time the team could show what it got wrong each week and who checked it.

How to assess your team

The free Support Maturity Check places a team on each lens in about ninety seconds. It asks eight questions and returns a stage for each lens, the lens to move first, and whether AI is the next move. It runs in the browser and stores nothing.

A full assessment goes further. Manic Edge interviews the people who run and work in support, reviews the artefacts they use, such as macros, QA forms and dashboards, and looks at ticket data where it exists. It ends with a scorecard of current and target stages across the six lenses and a written blueprint for the climb.

Take the Support Maturity Check About support strategy