Bookbag

AI engineering consultancy

We set up AI engineering inside your company.

Bookbag works with CTOs, CIOs and heads of engineering. We assess how your teams build software today, set up AI-assisted workflows in your own repositories and tools, train your engineers, put the controls in place, and build alongside your team until it runs without us.

Exhibit 1From licences to measured value
  1. 01

    Licences

    Coding assistants bought for every engineer.

  2. 02

    Experiments

    Teams try agents their own way. Results are anecdotes.

  3. The gap: tools without a model
    03

    Operating model

    Roles, review, workflows and budgets designed for AI-written code.

  4. 04

    Governed delivery

    Agents work the real backlog under policy and approval.

  5. 05

    Measured value

    Outcomes reported against a baseline the board accepts.

Source: Bookbag's engagement model. Our work covers stages 3 to 5.

What we sell
Consulting and hands-on engineering. We assess, set up, train, govern, build and run AI-assisted software engineering inside your company.
What we don't sell
Software licences, coding agents or outsourced developers. Your people and your tools stay yours, and you never have to buy software from us.

01/What we do

Every service we offer.

40 services in six groups, all in one specialism: software engineering. Each is work our people do inside your company, with your teams and your tools. Buy one on its own, or let an engagement combine them.

01

AI readiness assessment

We find out how your teams use AI today, what it costs, where the risks are, and which changes are worth making first.

AI engineering maturity review
Where each team stands, from first experiments to governed delivery.
Tool and licence audit
What has been bought, what is used, and what overlaps.
AI spend review
Licences, tokens and usage-based costs, and where they are heading.
Delivery baseline
Lead time, change failure, review time and throughput, measured before anything changes.
Risk and security review
Code, data and IP exposure from the way AI is used today.
Tool and vendor selection
Help choosing coding assistants and agents, judged on your codebase, not on demos.
Roadmap and investment case
What to change first, what it costs, and what it should return.

02

AI workflow setup

We set up AI-assisted development in your repositories, tickets and pipelines, with a person signing off where it matters.

Coding assistant rollout
Configured for your languages, repositories and security rules.
Agent workflows
From a ticket to a reviewed pull request: plan, code, test, review.
Codebase context
The rules, architecture notes and conventions agents read before they write.
AI code review
AI reviewers on every pull request, with a named person signing off.
Test generation
Agents that write and maintain tests, checked by your pipeline.
CI/CD integration
The checks AI-written changes must pass before they merge.
Ticket and backlog integration
Agents working from your tickets, in the tools you already use.

03

Engineering team enablement

We train engineers, reviewers and team leads to work with coding agents on their own codebase, not on a toy example.

Hands-on team workshops
On your own codebase, with your real tickets.
Engineer training
Context, prompting, testing, and when not to use an agent.
Reviewer training
How to review AI-written code quickly and safely.
Engineering leadership training
Running teams, estimates and hiring when agents do part of the work.
Playbooks and standards
Written practices your teams keep using after we leave.
AI champions programme
Engineers in each team who coach the rest.
Coaching and office hours
Help on hand while teams build the habit.

04

Governance & quality

We put the rules in place for AI-written code: who may use what, what needs approval, what it may spend, and how it is checked.

AI usage policy
What engineers may use, on which code and which data.
Agent access and permissions
Which agent may touch which repository, and with what rights.
Approval rules
Where a person must sign off, and who.
Spending controls
Budgets, limits and usage reporting for AI tools and models.
Secure setup
Secrets, data handling and licence checks for AI-written code.
Quality gates and evaluations
The tests and checks AI output must pass, and how it is scored.
Audit evidence
A record of what agents changed and who approved it, ready for security and compliance reviews.

05

AI-native software delivery

Need something built? We deliver it with AI-native engineering, in your repositories and alongside your team. Not outsourcing: your team sees and learns every step.

New features and applications
Built with AI-native engineering, in your repositories.
Legacy modernization
Framework upgrades, migrations and rewrites of ageing code.
Integrations and internal tools
Connecting your systems, and the tools your teams need.
Test coverage uplift
Bringing untested code under test before it changes.
Maintenance backlog
Bugs, upgrades and technical debt that never get prioritized.
Prototypes and proofs of concept
Fast, scoped builds to test an idea before you commit to it.

06

Managed AI engineering

After go-live we run and improve the setup with your team, until you are ready to take it over yourselves.

Workflow operations
Running and tuning your AI workflows as tools and models change.
Tool and model updates
New releases tested on your codebase before your teams adopt them.
Usage, cost and outcome reporting
Regular reports your leadership can act on.
New team onboarding
Bringing further teams onto the setup, one at a time.
Ongoing team support
A named contact for your engineers and leads.
Quarterly leadership review
What changed, what it cost, and what to do next.

02/How it works

Start small, prove it, then scale.

Engagements run in three steps, and each step ends with something your leadership team can decide on. Scope, team and fee are agreed in writing before we start.

Step 1

4 weeks

Readiness assessment

For a leadership team deciding where to start.

AI readiness assessment

What you have at the end

  • A baseline of AI use, spend and delivery performance today
  • Risk, security and governance gaps, ranked
  • A sequenced roadmap and an investment case

Flagship programme

12 weeks

AI Engineering Transformation

With two to four delivery teams, on their real backlog.

AI workflow setupEngineering team enablementGovernance & quality

  1. Weeks 1–2

    Baseline and design

    Agree the measures, the teams and the controls.

  2. Weeks 3–8

    Build and run

    Governed AI workflows on live work, reviewed weekly.

  3. Weeks 9–12

    Measure and hand over

    Evidence against the baseline, and a plan to scale.

What you have at the end

  • An operating model your teams run without us
  • Governed AI workflows in production on real work
  • Policies, approvals and budgets in force
  • An evidence pack measured against week one

Step 3

Ongoing

Scale and operate

For organizations taking the model to more teams.

Managed AI engineeringAI-native software delivery

What you have at the end

  • Roll-out to further teams and business units
  • Managed operations, if you want them
  • A quarterly review with your leadership team

03/Point of view

The tools are the easy part.

Almost every engineering organization now has AI coding tools. Far fewer can say what changed: whether delivery is faster, whether quality held, what it costs, and who approved what the agents wrote. That gap is an operating-model problem, not a tooling one, and it is the problem we are hired to close.

  1. 01

    Adoption is not the outcome.

    Usage dashboards measure activity. Leaders need lead time, change failure, review load and cost per change, measured against where they started.

  2. 02

    Review is the new constraint.

    When agents write more changes than people can review, throughput moves to the review queue and risk moves with it. The answer is policy, routing and evidence, not more reviewers.

  3. 03

    Control comes before scale.

    Policies, approvals, access and budgets are cheap to set before agents spread across teams. Retrofitted after an incident, they are not.

  4. 04

    Prove it in your own numbers.

    Vendor benchmarks do not transfer to your codebase. A programme earns its next budget with evidence from your own repositories.

Exhibit 2Adoption is nearly universal. Control and cost are not settled.

90%

of the technology professionals DORA surveyed use AI at work. The same study still finds a negative relationship between AI adoption and software delivery stability.

Google Cloud, Announcing the 2025 DORA Report, 23 September 2025

87%

of developers say AI helps them ship work faster. Only 28% say their day-to-day work has noticeably less friction than before AI.

Atlassian, State of Developer Experience Report, 2026 edition

49%

of developers spend four or more hours a week reviewing AI output, correcting prompts and similar work.

Atlassian, State of Developer Experience Report, 2026 edition

17%

of developers, and 23% of engineering leaders, say their organization has a clearly communicated, actively implemented AI strategy across the software lifecycle.

Atlassian, State of Developer Experience Report, 2026 edition

65%

of engineering teams using agentic coding will treat the IDE as optional by 2027, Gartner predicts, moving control, governance and validation to automated platforms.

Gartner, press release on enterprise AI coding agents, 20 May 2026

2028

is the year Gartner predicts AI coding costs will overtake the average developer's salary, as token consumption grows and vendors move to consumption pricing.

Gartner, press release on AI coding costs, 24 June 2026

Third-party research, paraphrased from each publisher's own release; the predictions are the publishers'. Atlassian's figures come from its survey of 1,500+ software practitioners. Bookbag has no affiliation with Google Cloud, Atlassian or Gartner.

04/How we work

What you can hold us to.

One specialism.
Software engineering is the only function we work on. We do not sell general AI strategy, data platforms or chatbots.
No software to buy.
We work with the repositories, coding assistants and models your teams already use. You never have to buy software from us.
Your backlog, not a demo.
Work starts on real items in your repositories within the first fortnight, under your review rules.
A baseline before anything changes.
We agree the measures in week one and report against them, including the ones that do not move.
Controls before scale.
Policies, approvals and budgets are in force before another team joins.
Senior people on the work.
The people who scope an engagement are the people who deliver it.

05/Research

Research and briefings for engineering leaders.

We would rather open a conversation by sharing what we have learned than by pitching. The study and the briefings are how we do that.

Study · first edition

The State of AI Engineering in Enterprise Software Teams

How engineering organizations are adopting AI-assisted development: what they use, how they govern it, what it costs, and what they can show for it. We are inviting CTOs, CIOs and heads of engineering to take part.

  • A short, structured interview
  • Findings shared with participants before publication
  • Your position against an anonymized peer group
  • Nothing attributed to you without written consent

Executive briefing

A working session for your leadership team

An hour with your CTO, CIO or engineering leadership on what is changing in AI-assisted engineering, what the public evidence says, and what it would take in your organization.

  • What the public evidence does and does not show
  • The operating-model choices in front of you
  • The first decisions to take, and what to measure

Not ready for a conversation? Score your organization first: 26 yes-or-no questions, about ten minutes, nothing to sign up for.

Use the readiness checklist

Start here

Start with a conversation, not a proposal.

Tell us where your engineering organization is with AI and what the board is asking. We will come back with a time for a briefing with your leadership team.