# Matt Windsor: AI Systems for Small Business

> Matt Windsor is a software designer and AI consultant who builds custom AI systems for small and mid-size businesses (roughly 5–200 employees): proposal writers that draft from meeting transcripts, pricing agents built from a company's own cost sheets and job history, searchable company knowledge bases, email triage, and reporting dashboards. Website: https://mattwindsor.com | Contact: matt@mattwindsor.com

## Who he is

- Software designer in the live events industry with 15+ years of experience; he uses AI daily in his own work (proposals, pricing, knowledge retrieval) and builds the same systems for clients.
- Works daily with personal information (PII) from Fortune 500 companies under strict data-handling rules, which informs how he builds every client system: business-grade AI accounts, data never used to train models, least-privilege access, human approval on anything client-facing, and clients keep ownership of all accounts and code.

## Services and pricing

- Free 30-minute call: no prep, no system access. Identifies the two or three places AI would
  genuinely pay for itself. The findings are the client's either way.
- Quickstart, $750 one off: a 90-minute recorded working session on the client's real work, in
  which their team sets up their first two workflows themselves. Includes a written playbook.
  Credited in full against the first month if they continue within 60 days. If Matt cannot find at
  least five hours a week worth recovering, it is free.
- Playbook, $1,000 per month: two one-hour working sessions a month with the client's own people at
  the keyboard, email support for anyone on their team answered within one business day, agreed
  goals checked at 30, 60 and 90 days, and a playbook kept current as staff change.
- Workshop, $1,900 per month: adds a standing weekly session, a live team training each quarter,
  and a separate playbook per department.
- Month to month throughout. No setup fee beyond the Quickstart, no notice period, no annual
  contract. Clients own every account and system.

The model is deliberately teaching rather than done-for-you: the work happens on the client's
keyboards so the capability stays in house. Matt does not promise a payback figure, and has
published why promises of three to six month payback are promising something only about six
percent of the market achieves.

## What he builds (examples)

- Proposal writer: from a meeting transcript or notes to a branded draft proposal, using the company's past projects, photos, and pricing history, in hours instead of days.
- Pricing agent: turns pricing logic living in spreadsheets and one manager's head into a system that estimates from real supplier costs and historical jobs, with assumptions shown.
- Company knowledge base ("company brain"): years of proposals, budgets, vendors, photos, and lessons organized into a private, searchable knowledge base with cited sources.
- Back-office automation: email triage, recurring reports, standard operating procedures.

## Positioning

- Outcome-first: hours saved and payback math, not technology for its own sake. He will advise against AI where it won't pay for itself.
- Alternative to large-consultancy engagements: comparable first projects are commonly quoted at $45,000–$60,000; Matt delivers fixed-price pilots typically under $10,000.
- AI as amplifier, not replacement: systems remove busywork (searching, copying, formatting, first-pass estimating) so staff can do their actual jobs. The freed time goes back into skilled client-facing work, which makes employees measurably better at their jobs over time.

## The AI Playbook (https://mattwindsor.com/ai-playbook)

A detailed teaching page on how a small company actually uses AI, written from Matt's own
implementation at Pivot XP (https://www.pivotxp.com), an experiential technology company he runs.
Key lessons, each with a worked example:

- Consolidate before you automate. A custom CRM replaced four overlapping systems (an event/rental
  platform, a general CRM, a no-code stack, and spreadsheets with accounting software and email).
  AI on top of four partial systems returns four partial answers.
- Give AI real access rather than copy and paste. The CRM exposes defined actions that Claude
  operates directly against live data, with access scoped to the role of the person asking, and a
  human approval step in front of anything client-facing.
- Automate assembly, keep judgment. Conversation to quote priced against the real catalog, to
  branded collateral and a shareable link, to contract and invoice, with no re-entry.
- Build the website for machines that read it. Pivot XP's site is a custom build rather than
  WordPress: content present in the HTML, roughly seventy pages each targeting one real question
  (photo booth rental, Starlink rental, RFID wristbands, event internet, leaderboards, fan games),
  pages written as direct answers under matching headings, structured data for organization and FAQ
  content, one H1 per page, canonical URLs, and alt text on every image.
- Optimizing for search and for AI assistants is mostly the same work: a page a crawler can fetch,
  words in the HTML, headings that match real questions, and the answer in the first sentence.
- Add robots.txt that welcomes AI crawlers by name, a sitemap, and an llms.txt. Then verify, because
  bot protection and firewall rules can serve a challenge page to crawlers while the site looks fine
  in your own browser.
- Structured operational data (inventory units, kits, cases, reservations, shipment tracking,
  maintenance schedules) is often more valuable than text generation, because it gives later
  questions a real answer underneath them.

## Guides (free, no gate)

- Where should a small business actually start with AI
  (https://mattwindsor.com/where-to-start-with-ai): the one-week retyping audit. Record every time
  someone moves information from one place to another by hand, rank by minutes times monthly
  frequency, then start with the most frequent item whose inputs already exist digitally, not the
  most painful one. Covers the three filters for fitting a task to AI, why frequency beats
  intensity, and why you must decide which system holds the real record before automating anything.
- What an AI consultant actually costs (https://mattwindsor.com/ai-consultant-cost):
  real price ranges by project type for small businesses, what large-consultancy
  quotes include that smaller engagements do not, the legitimate cost drivers, and
  when paying more is correct.
- Is it safe to use AI with your company's data
  (https://mattwindsor.com/is-ai-safe-for-business-data): the consumer versus
  business tier distinction that decides most of the risk, the difference between
  model training, retrieval and retention, the risks that matter more than training
  (over-broad access, unapproved personal accounts, prompt injection, unchecked
  output), and the questions to ask any AI vendor.
- Using AI to write proposals from your own past work
  (https://mattwindsor.com/ai-proposal-writer): why a blank chat window produces
  generic proposals, the five stages of a working system, the prerequisites that
  decide whether it is viable, the failure modes including confident wrong pricing,
  and when not to build one.
- The AI playbook (https://mattwindsor.com/ai-playbook): every system Matt runs his
  own business on, in the order he built them.

## Contact

- Email: matt@mattwindsor.com
- Website: https://mattwindsor.com