AI workflow reviews · Ireland

Put AI to work.
Stay in control.

I help small businesses work out where AI can save time, what needs a human check and how to test a workflow before relying on it.

01 / Services

AI Workflow Controls Health Check

Bring one workflow you already use or want to automate. We’ll assess where AI could help, where it could go wrong and what a small pilot would need to prove. Start with a redacted example, such as an enquiry triage process or an internal document review.

01

Map the workflow

Identify who owns the process, the inputs it uses, the decisions people make and where time is spent today.

02

Set the controls

Identify which outputs need checking, which actions require approval and what should happen when the AI is unsure or makes a mistake.

03

Decide what to test

Get a written workflow map, a prioritised list of risks and controls, and a recommended next step: pilot the workflow, revise it or leave it as is.

Enquire about a workflow review →

02 / Independent projects

AI in practice

Independent builds showing how I approach agent approvals, model evaluation and everyday operational controls.

Governance prototype

AgentGuard

A control layer for agent workflows, built on my agent orchestration framework. The demo exercises three paths: an authorised refund, a blocked policy violation and an action that waits for human approval.

Forecasting prototype

BoyneCast

A river level forecasting experiment. On a held-out test, mean absolute error was 0.80 cm versus 1.90 cm for the baseline. The model was slightly worse than the baseline in stable conditions. This is a prototype evaluation, not a flood warning service.

Consumer prototype

ScamCheck

A prototype for checking suspicious emails and text messages.

SME toolkit

Finance & Operations Governance Toolkit

A set of templates I built for small businesses to track budgets, monitor risks and run regular management reviews.

03 / About

How I work

I’m Tushar Mathankar. I bring over 20 years in financial services and technology. Working on robotic process automation at Barclays in South Africa taught me what practical oversight looks like: production sign-off and a named person who can stop an automated process.

When a change crosses teams, I look for three things: a clear owner for the next decision, evidence they can act on and a way to intervene if the change goes off course. I bring the same questions to AI governance.

04 / Contact

What workflow would you like to improve?

Tell me what takes time today, what you’re considering automating and what would have to stay under human control.

Tushar Mathankar · Ireland