AI integration consulting
Get your systems ready to put AI to work.
I help technical teams plan how to build AI into their own systems. The work is a final report, scheduled meetings where I review your systems and give feedback, or both.
Why now
AI models keep improving, and using them well depends on having your data and systems ready for them to act on.
That comes down to two things.
- You need control of how your data is structured and where it is stored.
- You need to build the tools that give AI agents secure access to that data.
With both in place, you can shape the responses and actions your AI agents take to match your proprietary process.
If you lead a team that is building a system like this and isn't sure which path to take, get in touch. I can help you plan the short and long term milestones and review your systems as you go.
What I review
We start by agreeing to the scope: which systems, and which questions you want answered. Then I work through five areas with your team.
- 01
Systems
The tools and databases you run, how they connect today, and which ones an AI tool would need to reach.
- 02
Data
Where the records live, who is responsible for each source, how current each one is, and what a model would see if it read it.
- 03
Access
Who and what can read or change each system today, including service accounts and shared logins.
- 04
Where AI may act
For each place AI could help, whether it should only read, whether it can draft, and what needs a person to approve before anything runs.
- 05
Model terms
Whether the provider accounts, settings and contracts you use keep your data out of model training.
What you get
A final report. It covers each area above, sets out the milestones in order, short and long term, lists the gaps to close first, and we walk through it together.
Review meetings. Where it helps, we meet on a schedule and I go through your systems and your plan with you, and give feedback.
The work can be the report, the meetings, or both. The report is yours. Use it with your own team or with anyone else you choose.
- 1Systems and data in scope
- 2Access as it stands today
- 3Where AI may act, and where a person approves
- 4Model terms for your data
- 5Milestones and gaps, in order
More than ten years writing code
I've worked with AI tools since ChatGPT launched in late 2022.
I designed, built and deployed a custom, AI-integrated software system, and I maintain it as its only developer. It has about 170,000 lines of application code, an automated test suite of about 166,000 lines, and more than 700,000 lines of backend and data-processing code.
My work with AI goes past prompting a model. I build the workflow around it: how the model reaches the data, what it can do with that data, and how it responds, so the results fit a defined process. That draws on my software development experience and my background in documenting and organizing data. A person approves the steps that matter, and I keep refining the workflow as the tools change.
What it does not include
The work is advice, review and planning. Nothing gets built or installed. It is not a penetration test or a security certification, and I don't guarantee an outcome.
Underlay is my practice. It builds custom software for companies that want it built for them. The work on this page is for teams that are building it themselves and want to know what it takes to succeed with AI in their own organization. If you want a system built for you, see Underlay.
Get in touch
Email me the systems you're working with and what you want AI to do for your team. I'll write back with whether I can help and what a first call would cover.
Writing to me doesn't commit you to anything.