Damian Mathews

01 The pattern

Most companies bolted AI onto a few jobs and stopped there

Every business I've worked inside has the same shape. A handful of people use ChatGPT to draft emails. A team or two has a Copilot license. Maybe someone built a chatbot. Then the work plateaus, because the people who can see what to build next are not the people doing the day-to-day work.

The interesting part of AI in 2026 is not the models. It's the connective tissue. Pulling data from one system, drafting in another, summarizing a meeting, routing a decision, watching for an event, kicking off a workflow. Most of the repetitive, boring, or costly work your team does in a week is a candidate for this. The job is to find which parts, build them carefully, and bring the team along.

Discovery

I learn how your business runs by watching it run. Sit in on calls, shadow a few roles, look at your tools and data, talk to the people doing the work. The output is a written map of where AI fits and what to build first.

Build

I build the system. Real working tools that plug into the work your team already does. Not a demo. Not a slide. Something that gets used on Monday morning and saves time by Friday.

Training

I sit with the people who'll use it. Show them how, write down what I built, and stay in the room until they're confident. Adoption fails because nobody trains the team. I don't make that mistake.

Compound

Once one workflow is running on its own, the next is obvious. Most engagements grow into a roadmap. The end state is a business where the boring middle takes care of itself and your team focuses on work that needs a human.

02 What I do

Three ways an engagement starts

Pricing depends on scope, so we figure that out on a call. Every engagement starts with a free intro conversation. If we're a fit, we talk numbers. If we're not, I'll usually point you to someone who is.

/ 01

The Audit

Two to four weeks inside the business. I learn how the work flows, where the time goes, what your stack looks like, and what your team is already doing with AI. The deliverable is a written report: where AI fits, what to build first, what to leave alone, and what it'll take. You can run with it or roll into a build.

Process mapping AI fit assessment Roadmap Stack review

/ 02

The Build

Six to twelve weeks. I build what the audit pointed at: an internal assistant trained on your data, a research and drafting workflow, a sales call summarizer, a document processing pipeline, an automation between three tools nobody wants to wire up. I write the prompts, build the integrations, document the system, and train the team that'll own it. You get working software, not a deck.

Internal AI tools Workflow automation Custom integrations Documentation Training

/ 03

Embedded

A few hours a week, ongoing. I maintain the systems we built, add new ones as they come up, train new hires, and keep an eye on what's changing in the AI landscape that you should care about. The cheapest version of having a senior AI person on the team, without the full-time hire. Most builds graduate into this.

Ongoing iteration Maintenance Team enablement New builds Strategy

03 The engagement

How a typical project moves

Every engagement is shaped to the business, but the rough arc is the same. Five phases. Some get long, some are short. The whole thing usually lands somewhere between six and sixteen weeks for the first round, then settles into the embedded rhythm.

Phase 1

Discovery

Sit in. Shadow. Watch the work. Ask questions. Read your docs. Look at your stack. By the end I know how things actually run.

Phase 2

Map

Written report of where AI fits. Each opportunity gets a priority, an effort estimate, and a sketch of what the build looks like.

Phase 3

Build

Pick the highest-leverage thing. Build it. Test with real users. Iterate. Ship something that gets used Monday and saves time by Friday.

Phase 4

Hand off

Train the team. Write the docs. Make sure the people who'll own the thing can run it without me. Set up metrics so we can see if it's working.

Ongoing

Compound

Embedded retainer. New builds, maintenance, new hires trained, watching for what's changing in AI that you should care about.

04 What gets built

The kinds of things I build

No two projects are the same, but the patterns repeat. Each of these has shipped inside a real engagement. The shape changes, the underlying work doesn't.

Document drafting systems

Long-form documents drafted from your inputs in your voice. Team reviews and finalizes; hours collapse to minutes.

Internal AI assistants

Trained on your documents, your data, your way of doing things. Available in Slack, Teams, or a custom interface.

Meeting and call workflows

Transcripts cleaned and summarized, action items pulled and pushed into your CRM or task tracker. The post-meeting cleanup nobody wants to do.

Cross-tool automation

Stitching your stack together. An email triggers a CRM update, which triggers a draft, which lands in your inbox for review. No more copy-paste tax.

Reporting automation

Weekly reports, dashboards, and status summaries pulled from multiple sources, written in plain English, in your inbox before Monday standup.

Research pipelines

Pull data from the web, internal docs, broker research, and transcripts; synthesize and output a structured first pass your analysts can finish.

Onboarding and training

Get your team using AI tools well, new hires up to speed faster, fewer support tickets. The unsexy work that makes everything else stick.

Custom internal tools

The thing your team has been asking for that no SaaS vendor sells. Built fast, hosted simply, owned by you.

05 The arc

Every workflow lives somewhere on this line

Most processes start manual. Then someone uses ChatGPT for a piece of it. Then a real tool gets built. Then the tool runs on its own. The end goal isn't replacing your team. It's freeing them from the work that doesn't need them.

From manual to agentic. Drag to explore.

Interactive · 4 stages

Stage 01

Manual

Stage 02

Assisted

Stage 03

Augmented

Stage 04

Agentic

Stage 01 · Manual

A team member reads the documents, takes notes, drafts the output by hand. It works. It's slow.

Most companies start here. The work is real, the people are real, but most of the keystrokes are mechanical. The slow part isn't the thinking, it's the typing, the copying, the formatting, the cleanup.

06 The stack

What I work with, day to day

Tool-agnostic but opinionated. I'll work with whatever your business already runs on, and recommend additions when they earn their place. The list below is what I reach for most often. The right tool depends on the job.

Frontier model

Claude

Anthropic · primary

Agentic

Claude Code

Build pipelines

Workspace

Claude Cowork

Team AI workspace

Frontier model

GPT

OpenAI · when fit

Agentic

Agent Builder

OpenAI platform

Protocol

MCP

Tool integrations

Hosting

Vercel

Apps and APIs

Development

React, Next.js

Custom apps

Data

Python, SQL

Glue and pipelines

CRM

Dynamics 365

Microsoft stack

Workspace

Microsoft 365

Teams, SharePoint

Workspace

Google Workspace

Drive, Gmail

07 Who's behind this

I started my first business, Newton Kicks, during the 2020 lockdowns. While still in online school, I generated over $500k in revenue in the first 18 months. That's where I learned digital marketing, social media strategy, sales, and what it actually takes to grow a company from zero.

Then ChatGPT launched and I was hooked from day one. I've spent every spare hour since learning the ins and outs of prompting, building automations, vibe-coding apps end-to-end, and figuring out how to make AI earn its keep inside real businesses.

Today I bring both halves together: a founder's instinct for what businesses actually need, and a builder's toolkit for shipping it. I've worked with asset managers, multi-location operators, marketing teams, and founders who needed something built yesterday. Different industries, same shape every time: find the bottleneck, build the thing, hand it back.

Most of the value lives in the boring middle of a company, where systems meet people and someone has to translate between them. That's where I work.

Based

Boston, MA

Engagements

Audit · Build · Embedded

Reachable

Email · Calendly · LinkedIn

08 Common questions

Things I get asked

Does my team need to be technical to work with you?

No. Most of the people I work with are operators, executives, analysts, and team leads. The point of bringing me in is so your team doesn't have to figure out the technical part. I do that. Your team's job is to tell me what their week looks like, what's annoying, and what they wish would just happen on its own. I take it from there.

How do you handle sensitive data?

Carefully. I work under NDA on every engagement. I use enterprise-grade plans that contractually exclude your data from model training. I'll redact identifying information before anything touches an AI tool when the situation calls for it, and for the strictest cases, we deploy on infrastructure you control. If your industry has specific compliance requirements (financial services, healthcare, legal), we scope around them on the first call.

What size company do you work with?

Mostly between 10 and 500 employees. Small enough that decisions move and the people doing the work talk to me directly. Big enough that there's real complexity to work with. I've also done one-off projects for larger orgs and for solo operators when the project was the right shape. Size matters less than whether the work is buildable and the buyer is serious.

How long is a typical engagement?

An audit is two to four weeks. A build is six to twelve weeks for the first deliverable. Embedded engagements are ongoing, month to month, no long-term contract. Most clients start with a build and graduate into embedded once the first system is running. The point is to keep showing up as long as I'm earning my keep.

What does it cost?

Depends on scope. The intro call is free, and by the end of it I usually have enough to give you a real number. I don't publish prices because the variation is huge. Some audits are a few thousand dollars and some builds run into five figures over a few months. The right way to find out is a 30-minute call.

Do you build with one specific AI tool, or do you pick?

I pick. Claude is what I reach for most often, but I use GPT, Gemini, and open-source models when they're the right fit for the job. The best tool depends on the work, the data, the budget, and what your team is already paying for. Part of what I bring is knowing which one to reach for and why.

Can my team maintain what you build after you leave?

That's the goal. Every build comes with documentation written for whoever's going to own it, plus a training walkthrough. Most clients keep me on an embedded retainer because they want me around to add new pieces and watch for what's changing in AI. The system is yours either way.

Do you work with companies outside Boston?

Yes. Most of my clients are remote. I'll fly in for kickoffs, key reviews, and training sessions when it matters. Discovery and build phases are mostly remote with regular video time. Geography hasn't been a constraint in years.