AI strategy and advisory
Aadhib Nasser · AI Engineer and AI Consultant
Who this is for
- 01Leadership teams that need a few AI decisions made well, rather than a strategy document
- 02Companies choosing between models, vendors, or building and buying
- 03Anyone about to start a pilot who wants its success criteria written down first
What the work covers
- 01Finding the workflows where AI would change the outcome, and the ones where it would not
- 02Model and vendor selection, and re-evaluation when prices or models change
- 03Build, buy or configure decisions
- 04Pilot design, with success criteria written before the pilot starts
- 05Deployment options: hosted, local or hybrid, and what each asks of your team
- 06Vendor and platform risk: deprecation policies, second providers and exit paths
How I approach it
Decide before the pilot
A pilot that ends in "it worked" has not decided anything. The decision it feeds, and the evidence that decision needs, are agreed first.
Your tasks, not the launch table
Vendor benchmarks are a reason to test, not a result. Models are compared on a representative set of your own tasks, with a written pass criterion and the cost per completed task.
Plan for the model to disappear
A deprecation policy promises notice, not a successor. Every capability you depend on gets an interface you own and a tested second provider.
Related work
Case studies and projects
Further reading
Notes on this subject
Questions
Asked before an engagement
Where should a company start with AI?
With one workflow where the cost of the current process is known and the output can be checked. That gives the pilot a baseline to beat and a way to tell whether it did.
How do you choose between models?
On your own tasks. A representative set pulled from real work, a written pass criterion for each, and the cost per completed task including retries. Then the decision is written down: which task types go to which model, and why.
Start a conversation
I usually reply within a day
Send a few sentences about the problem: what the process is today, where the data lives, and what would change if it worked.