Two categories of enterprise AI work.
The first is theatre: chatbots, demos, internal showcases. It produces slides, not throughput. The second is operational: agents that remove work from a person's day, workflows that decide cases that used to escalate, reports that generate themselves overnight.
We do the second, and the success criterion is always the same: how many hours of operational work have been removed, and what the team is doing with that time now. If nobody can answer the second half, the first half was theatre.
Why enterprise AI stalls.
Level four applied to a level-two problem
A deterministic rule would have solved it. A model was used instead, and now the decision cannot be explained to an auditor.
No baseline
Nobody measured how long the task took before, so no one can say whether anything improved.
Data not ready
The use case depends on master data that is duplicated, incomplete, and owned by nobody.
Nobody will sign off on it
An operations director will not let software decide anything it cannot explain, bound or replace.
A pilot that cannot scale
Built outside the operating systems, it works in a demo and cannot be put into production.
Hours removed, then reabsorbed
Time is freed and immediately consumed by other unmeasured work, so the benefit never appears.
What the engagement actually includes.
Opportunity assessment
Half a day to two weeks reviewing workflows and naming which can lose hours this quarter — and which cannot, and why. The output is a ranked list with an estimate against each.
Automation level decision
Each candidate is placed on the maturity ladder — workflow, rules, AI-assisted or agentic — and we choose the lowest level that solves it, because each level down is cheaper and more governable.
Baseline measurement
Time the work before anything is built. Without a measured starting point, no hour-removal claim afterwards is credible.
Governance model
Authority limits, escalation rules, audit logging, model change control and the human review points — designed before deployment, not retrofitted after an incident.
Operating model for AI
Who owns the models, who approves a change, who is accountable for a decision the software made, and how a new use case gets from idea to production.
Roadmap
Sequenced so the first delivery is small, provable and in production within a quarter — not a two-year programme with a demo at the end.
From workflow review to production.
We are not locked to one model or provider, and the replacement path is defined before deployment. These are the components an AI programme typically touches.
Ranges observed on Al Jawad engagements. Your targets are agreed in assessment, before the work starts.
What does the consultation cost?
The half-day AI opportunity consultation is an advisory conversation, not a sales call. We review your workflows and name which can lose hours this quarter — and which cannot.
Do we need our data cleaned first?
For some use cases yes, for others no. Part of the assessment is separating candidates that need a data programme first from ones that can proceed now.
Will you recommend against AI?
Often. Most candidates we review are better solved by a workflow or a deterministic rule, and we say so — those answers are cheaper, more reliable and easier to govern.
Which models do you use?
Whatever fits the constraint — including none. We do not architect lock-in to a single model or provider, and the replacement path is defined before deployment.
Start a conversation.
Choose the one that fits where you are. None of them is a sales call. Each is an advisory conversation calibrated to a specific question.