Last Mile AI - Ireland

The gap between
AI potential
and AI results.

Most organisations know there's an opportunity. The hard part is the last mile - getting from possibility to something that actually works in your business. That's what I do.

Start a conversation

Selected outcomes

€3.5M
Waste reduction per plant, manufacturing client
6wk
Work saved annually, professional services tool
15+
Years delivering decision-critical systems
4G
National rollout optimised, beating McKinsey

Selected work

What it looks like in practice

Across manufacturing, financial services, telecoms, professional services, and enterprise supply chain. Each project built around a specific business problem - not a technology showcase.

Intake and triage

Government, financial services

AI that reads, sorts and drafts replies to everything coming in the door

Shared inboxes nobody owns, routine questions answered by hand, jobs raised with half the details missing. We build AI that sorts incoming email, forms and documents by your rules, drafts the routine replies and hands a complete case to the right person. Money, complaints and anything unusual always go to a person first, and every correction is logged so you know the real error rate. The full use case
Inbox
sorted, drafted
and handed over
Workflow systems

Built for ourselves first

One place for every job, instead of Excel and email

When each job lives in a spreadsheet, three email threads and someone's head, nobody knows what stage it's at. We map how the work really moves, then build the smallest system that fits: one record per job, reminders and approvals in the system, reports that come out of the work as it's done. Built AI-ready, so agents can work in it later. We run our own business on one. The full use case
1
record per job,
not five
Data and reporting

Retail bank, food producer

One view of the business, from systems that don't talk to each other

Sales in one system, stock in another, finance in a third, and the real numbers in a spreadsheet. We connect what you have, agree one definition of each number and put dashboards in front of the people who use them. Sized to the business: a tidy database and a few dashboards for a small firm, a governed Databricks platform for a retail bank. The full use case
1
version of
the numbers
Training and enablement

Business owners and teams

Everyone using AI properly, with rules people understand

Most teams are already using ChatGPT or Claude, each in their own way, with no rules about what goes in. We run hands-on workshops for owners, managers and staff, from the fundamentals to building a first agent live in the room and then trying to break it. People leave with clear rules on what AI may do on its own and when it has to ask, and the confidence to use it on their own work. Ask about a workshop
Hands-on
workshops for
non-technical teams
Manufacturing

Global cocoa manufacturer

Eliminating raw material waste across complex production constraints

A large manufacturer was losing significant value to raw material waste that was difficult to control across multiple competing production variables. We designed and built a mathematical optimisation system - a mixed-integer linear programme - that balanced all constraints simultaneously and recommended production decisions in real time. Built from scratch to full production deployment.
€3.5M
annual savings
per plant
Telecoms Infrastructure

4G national rollout, Indonesia

Sequencing a national 4G network rollout across thousands of sites

A telecoms infrastructure client needed to roll out 4G coverage across Indonesia - a highly complex sequencing problem with geographic, commercial, and technical constraints. We built an optimisation programme that planned the rollout efficiently. Won the contract competitively - beat McKinsey in the bid. Delivered the programme on time.
4G
national rollout
delivered
Enterprise SaaS

Stanley - supply chain forecasting

Full multi-tenant forecasting platform embedded into supply chain operations

End-to-end design and build of a multi-tenant SaaS forecasting platform for a major industrial client. Complex analytical models embedded directly into operational workflows - FMCG supply chain, manufacturing, spare parts planning. CFO-level stakeholders. This was not a prototype handed to an internal team; it went into production as a working operational system.
SaaS
full product
end-to-end
Financial Services

Regulated financial firm

Automating suspicious transaction review - and learning from the reviewers

A compliance team was spending hours each day manually reviewing transaction reports for suspicious activity. Rather than build a rule-based system, we worked closely with the reviewers to understand how they actually made decisions - then built a machine learning model that learned directly from their judgement. The model automated triage and prioritised cases by risk level, giving the team back significant time while maintaining the quality of their oversight.
hrs/day
manual review
eliminated
Professional Services

Mortgage brokerage

Turning a good idea into a production tool - in a week

A mortgage broker built an AI tool to generate structured client profiles. The logic worked. What they couldn't figure out was how to share it securely with colleagues and keep it running reliably. A few days of work got it into production - secure, shareable across a small team, and running continuously. The broker calculated it saves around six weeks of work per year across the team. This is the last mile.
6wk
saved annually
per team

Who I work with

Three kinds of engagement

Different organisations need different things. Here's where I typically add the most value - and what that looks like in practice.

01

SMEs & owner-managed businesses

You know there's something here. You're just not sure where to start.

AI has opened up capabilities that used to require large technical teams and big budgets. I help you find the right unlocks, prioritise them, and deliver the first one quickly.

How this works →

Tech-forward companies

02

You've got technical capability. You need someone who understands the domain as well as the stack.

The most common failure in AI delivery isn't technical - it's the gap between what gets built and what the business actually needs. I get deep into your domain fast and deliver production-grade systems end to end.

See the work →

Fractional CTO / CDO placement

03

You need senior technical leadership, embedded - not just advisory.

Available for fractional engagements from financial advisory through to hands-on delivery leadership - at board level, leading delivery teams, or both.

Talk to me →

How it works

From first conversation to working system

No long strategy phases before anything gets built. A first conversation is informal and takes about 30 minutes.

01 -

Diagnose

A short discovery session to identify and rank AI opportunities in your business - what's high value, quick to deliver, and the right place to start.

02 -

Build

One well-defined pilot, typically 30–90 days, tied to a specific business outcome. Built and delivered into production - inside your actual workflow, not beside it.

03 -

Iterate

We measure results against what we set out to achieve, then decide what's next. Most first projects generate the ideas for the second.

Services

What I actually do

Often in combination on the same project.

  1. AI Strategy & Opportunity Design
  2. End-to-End AI Delivery
  3. AI Operating Model Design
  4. Embedded Decision Systems

About

15+ years. Multiple industries. One consistent edge.

I've spent 15 years delivering decision-critical systems in financial services, manufacturing, telecoms, and enterprise software. The work has ranged from mathematical optimisation in industrial production to machine learning in regulated financial workflows to full SaaS product delivery.

What connects it is domain immersion. Technical capability is table stakes. The differentiator is understanding a client's business well enough to build something that actually solves the right problem. I've worked closely enough with clients in shipping, supply chain, and financial compliance to anticipate requirements before they can fully articulate them - and the quality of what gets built reflects that.

Full lifecycle delivery Business requirements → architecture → build → production → ops. No hand-off to another team at the difficult part.
Strategy and execution in the same person Unusual enough that it generates scepticism. The case studies are the answer.
Cloud-native and on-premise AWS, Azure, GCP. Infra, DevOps, and security included - not subcontracted.
Modern AI tooling and classical methods LLMs, agentic systems, ML, mathematical optimisation. The right tool for the problem, not the fashionable one.

Get in touch

Let's talk

If you're serious about putting AI into day-to-day operations - not just proofs of concept - get in touch. A first conversation is informal and takes about 30 minutes.

info@lastmile-ai.com

No pitch deck required. Tell me what you're trying to solve.

Most projects are scoped and started within two weeks.

Location Ireland - working internationally
Typical first step 30-minute discovery call
First delivery Pilot project, 30–90 days
Engagement types Project, retained, fractional