AI in Customer Support: Decide Before You Deploy

I’ve built and scaled AI-powered customer support in real production environments globally (Agara → acquired by Coinbase).

Today, I work selectively with UK mid-market teams to help them make the right AI support decisions — before tools, pilots, or long programmes.

Start A Conversation

AI Customer Support Decisions

I work with teams who already know they want to use AI in customer support, but want to be confident they’re making the right calls before deploying tools or running pilots.

What to Automate vs Not

Identifying which customer interactions are genuinely suited to AI — and which ones are better left human-led to avoid brand, trust, or escalation issues.

Vendor & Approach Evaluation

Helping teams assess tools and approaches based on real-world constraints, not demos or marketing claims.

Pilot Design & Decision Criteria

Structuring small, controlled pilots with clear success and failure thresholds, so outcomes are unambiguous

How Is The Advisory Work Structured?

This work is designed to be short, focused, and decision-driven. The aim isn’t to explore AI broadly or run open-ended programmes, but to help teams make a small number of high-impact decisions clearly and with confidence.

Typical Structure

Short, fixed-scope engagement

The work runs as a defined sprint, usually over a few weeks, with a clear start and a clear end

Grounded in your real support context

Decisions are based on actual customer support data, workflows, and constraints — not abstract frameworks or generic benchmarks

Focused on decisions, not delivery

The output is clarity: what to do, what not to do, and how to proceed safely. I don’t implement tools or own execution

Designed to end with a go / no-go

The engagement concludes with a clear recommendation and decision criteria, so teams can move forward — or stop — without ambiguity

What You Get Out It

  • Clear agreement on what should and shouldn’t be automated

  • Reduced risk of choosing the wrong tools or approaches

  • A pilot structure that can be evaluated objectively

  • Confidence to proceed — or to stop — without second-guessing

What This Is Not

  • Tool implementation or configuration

  • Long-term programmes or retainers

  • Change management or training

  • Custom AI or software development

If this starts to look like a long internal initiative or delivery effort, it’s usually a sign to pause. This work is most effective when it stays short, contained, and decision-focused.

Who This Is For (and Who It Isn’t)

This is a good fit if:

  • You’re already serious about using AI in customer support, but want to be confident you’re making the right decisions before deploying tools or expanding pilots

  • You operate a mid-market business with meaningful customer support volume and real operational constraints

  • You want clarity and judgment, not a long programme, transformation roadmap, or vendor pitch

  • You’re comfortable with a short, fixed-scope engagement that ends with a clear decision

  • You value restraint — knowing where not to apply AI is as important as knowing where it can help

This is probably not a fit if:

  • You’re looking for someone to implement tools, manage vendors, or run delivery

  • You’re still in early exploration mode and just want to “learn about AI”

  • You want an open-ended engagement, ongoing retainer, or internal change programme

  • You want validation for a decision that’s already been made

  • You’re expecting AI to solve fundamental process or organisational issues

This work is designed to reduce risk and unnecessary complexity — not to push AI adoption for its own sake. If that framing resonates, we’re likely aligned.

Who Is This Coming From?

I’ve built and scaled AI-powered customer support systems in real production environments across the world. The products I have built have handled millions of emails, text messages and phone calls. The clients using my products have reduced costs by as much as 40% while maintaining service quality and CSAT.

I co-founded Agara, an AI customer support platform that was acquired by Coinbase, and later led product initiatives there focused on AI-driven support at global scale.

That experience shapes how I approach AI today: with a bias toward restraint, clarity, and decisions that hold up once systems are live and customers are involved.

What That Experience Includes

Building AI support products end-to-end

From early design through deployment in live, high-volume customer environments

Operating at Scale, Under Real Constraints

Working with support, operations, and compliance teams where mistakes affect trust, cost, and brand — not just metrics

Seeing Where AI Fails Quietly

Automation that looks good in demos but breaks down in edge cases, escalations, or customer sentiment

Balancing Automation with Human Judgment

Knowing when AI adds leverage — and when it creates more problems than it solves

This advisory work exists alongside building new products and is intentionally selective, short, and decision-focused. It’s not about selling tools or running programmes — it’s about making fewer, better calls before things are deployed.

Start A Conversation

If you’re already thinking seriously about using AI in customer support and want a grounded, decision-focused perspective before deploying tools or expanding pilots, feel free to reach out.

I’m selective about the work I take on, but I’m always open to an initial conversation to see if this is a good fit.

Get In Touch

Products I Am Working On

I spend some of my time building and shipping consumer products. My current focus is building a suite of fan engagement mobile apps for followers of Formula 1, tennis and pickleball.

Learn More