Conversational CodingA Backstory project Install the harness →
Menu

AI has made coding faster. Has it made your engineering better?

Spend five days with us and we’ll talk to the people doing the work, examine the agreed projects and repositories, then show you where AI is genuinely buying back time, where it’s quietly creating rework or risk, and what your team should do over the next 90 days.

What we’ll examineThe work around the code
5 focused days
How your people are actually using AIWhich work, which tools and who checks the result
Whether project knowledge survives the chatPurpose, decisions, constraints and history
What happens before and after generationPlanning, review, testing and release
Where quality or risk is movingSecurity, rework, operations and maintenance
Who can say ready and doneNamed human judgement and accountability
90 Days
A plan your team can actually followClear priorities for leaders, managers and the people doing the work.

You probably don’t need another AI maturity score.

You need to know whether the work is getting better, whether the people doing it understand what’s being shipped, and where the apparent speed is creating problems somewhere else.

The pull request arrived quicker. Did review, testing or rework get slower?

We follow real work from request to accepted change so you can see where AI saves time and where it merely hands the effort to somebody else.

Could every engineer explain the AI-generated code they committed last week?

We look at the repository evidence, the review and test discipline, the security checks and who owns the code once the conversation has ended.

If your most experienced AI user left, would the method leave with them?

We test whether purpose, decisions, constraints and lessons have become project knowledge, or whether they’re still trapped in chats and individual heads.

What will break when one successful pilot becomes several teams and projects?

We identify the practices worth sharing, the decisions that need named owners and the local context that shouldn’t be flattened into one corporate prompt.

Your engineering system
What you’re trying to achievePurpose, boundaries and the decisions AI mustn’t make
How people use AIWhat leaders expect and what engineers really do
What reaches productionReview, tests, security, release and support
What the project remembersDecisions and context that survive the chat
How work becomes accepted changeIntent, repositories, hand-offs and evidence

We look at the work around the code, because that’s where AI adoption succeeds or falls apart.

AI can produce a huge amount of plausible work before your team has agreed what good looks like. We look beyond the tools to how your team plans, builds, reviews and supports its software. You’ll see where AI is genuinely helping, where it’s creating more work, and what needs to change.

That’s why we examine the software delivery lifecycle, the pipeline around it, the quality of the code and evidence being produced, and the way your team captures and reuses what it has learned. Tool adoption on its own tells us very little.

We also look at where specialist perspectives could improve design, testing, review, security and stakeholder alignment. You won’t get a generic score with your logo pasted on it. You’ll get a prioritised account of what’s working, what needs attention and what’s worth trying next.

  • Conversations with executives, engineering managers and practitioners
  • Evidence from the projects and repositories covered by your assessment level
  • Recommendations tied to practical changes, named owners and the next 90 days

We use AI in the assessment. It doesn’t mark its own homework.

AI helps us work through more evidence consistently. It doesn’t understand your organisation, weigh the trade-offs or approve the conclusion. That remains human work.

01 / HUMAN DISCOVERYWe talk to your people.

We hear from the people setting direction, managing delivery and doing the engineering. The same practice can look very different from each seat.

02 / AI-ASSISTED SYNTHESISAI does more of the heavy lifting.

It helps us compare interviews, repositories and project evidence without losing the contradictions that deserve a closer look.

03 / HUMAN JUDGEMENTWe make the call and stand behind it.

We challenge the analysis, set the priorities and explain why each recommendation deserves time from your team.

You’ll be able to see what supports a conclusion, where uncertainty remains and who made the final judgement.

What might a useful finding look like?

Illustrative assessment output, not a client result.

Code is moving faster. Permission checks aren’t keeping up.

The sampled workflow has tests for a successful export, but none for a user trying to export another customer’s records. Reviewers have no agreed evidence requirement for that boundary.

Protect now
Make the ownership boundary explicit and add the missing negative test before the next release. Owner: engineering lead.
Improve next
Agree a review checklist for permission-sensitive changes and use it on the next three changes.
Check within 90 days
Sample those changes. Did reviewers see the relevant evidence before approval, and did any exceptions get a named decision?

The value is a specific change the team can make and a way to check whether it helped, not a score with no explanation.

The board, the engineering manager and the developer don’t need the same report.

90days of prioritised change, based on what we found rather than a generic transformation checklist.
For executives

A concise view of what AI is changing, the risks and opportunities that matter, and the decisions that need sponsorship or investment.

For engineering leaders

A practical account of the pipeline, hand-offs, quality, knowledge and standards that are helping the team or getting in its way.

For practitioners

Changes engineers can use in daily work, including where better context, review, tests or specialist challenge will make a difference.

For each sampled project

Specific findings from the project or repository, so an organisation-wide average can’t hide a weak delivery path or flatten a strong one.

For the next 90 days

A sequenced roadmap separating the things you should protect now, the improvements to make next and the experiments worth running before you commit more widely.

The price changes with how widely you want us to look.

Every level takes five days and gives you findings for executives, engineering leaders and practitioners, plus a 90-day roadmap. The difference is the number of people, teams and projects we can examine properly in that time.

One team or product

Focused

£4,000+ VAT

Five-day assessment

  • One sampled project or repository
  • Leadership and practitioner discovery
  • Executive, management and practitioner outputs
  • Prioritised 90-day roadmap
  • Complete persona library for 12 months, for up to 10 people
Ask about Focused →
Organisation-wide direction

Organisation

£15,000+ VAT

Five-day assessment

  • Three sampled projects or repositories
  • Executive, management and practitioner evidence
  • Systemic themes across teams and delivery paths
  • Organisation-level priorities and 90-day roadmap
  • Complete persona library for 12 months, for up to 100 people
Ask about Organisation →

VAT: All prices shown are exclusive of VAT. VAT will be added where applicable. Larger team licensing or a wider project sample can be scoped separately.

A report is useless if it lands in a folder and everyone goes back to the old way of working.

Every assessment includes a year of access to the complete persona library for the number of people shown in the pricing table. Your team can keep bringing engineering, architecture, testing, security, product, governance and leadership perspectives into the work while it puts the 90-day plan into practice.

A persona gives a general AI conversation a more disciplined point of view. It can challenge a weak assumption, spot a question the team hasn’t asked or help somebody prepare for a review. It doesn’t replace qualified advice, evidence checking or the person accountable for the decision.

294specialist personas
33collections
12months of access

For this to be useful, we need to see real work.

A prepared demonstration will only tell us how the process behaves on its best day.

  • Time with people from leadership, engineering management and daily delivery
  • Access to the project and repository evidence agreed for your assessment level
  • An honest account of what’s working, what’s awkward and where confidence is low
  • A named person who can challenge our interpretation before it becomes a recommendation

What this isn’t.

It isn’t an automated score, a certification or five days of sales theatre for a much larger engagement.

  • AI doesn’t approve the assessment or make the accountable decision
  • A positive finding isn’t a guarantee that future generated code is safe
  • The findings and 90-day plan are meant to be useful without buying another service
  • Where sensitive code can’t leave your environment, we’ll agree the evidence boundary before the work starts

Five days should leave you with answers, not another sales pitch.

Tell us what your team is building, how AI is involved and what you’re no longer sure you can trust. We’ll help you choose the breadth that fits, and if an assessment isn’t the right next step, we’ll say so.