Steve Taylor

Steve Taylor AI & Automation Lead

AI your whole team actually uses.

I build the agents and automations, then show everyone, technical or not, exactly what changed and how to use it. I am looking to do that full-time, on your team.

Full-time only

Talk about a role Read my resume
Exploded view of a sales automation pipeline with an AI agent at the center

My AI portfolio

8 builds. Compare the Majesty screenshots, or tap a numbered part of a demo.

Majesty Mortgage

Website redesign · Work in progress

Turn a complex subject into a website people can understand and use. I brought research, writing, design and development into one repeatable process, using AI to do the heavy lifting while I directed and reviewed the work.

Estimated manual effort
160–280hResearch and article writing alone
My direction and review
5hDelegating, reviewing and managing the work
Researched articles
120+A searchable Learning Center
What goes into the estimate

Manual effort is an estimate: a 40-hour research week plus 120 articles at 1–2 hours each. It excludes other page copy, design and development. The reusable process took roughly a month to establish on an earlier website; editorial review and refinements remain for this project.

Before & afterExisting website → Redesign preview
Dark-theme Majesty redesign homepage: guidance for a first home, a next move or refinancing, beside a house photograph. Existing Majesty homepage: Mortgage made easy, with prequalification and contact buttons over a photograph of a couple holding house keys.

Drag the divider or use this slider’s arrow keys. You can also choose either full view above.

Visit the existing websiteVisit the redesign preview

Learning CenterExplore the guides
Dark-theme Learning Center, with a guide search, topic filter and researched articles about preapproval, closing costs and mortgage documents.

Find a useful answer through search or topic filters. These researched articles are drafts awaiting editorial review.

Actual website captures from . The redesign and Learning Center are shown in their dark theme and remain works in progress.

How it works
Replaced
An outdated website that gave buyers little help understanding their options and few clear reasons to take the next step.
Built
A website organized around the buyer’s next move, with clearer guidance and useful next steps. The Learning Center brings 120+ researched mortgage articles into the experience, with search and topic filters.
Result
A working website preview and a substantial education library, with about five hours of my direction and review. This is still a work in progress: editorial review, bug fixes and refinements remain.
On your team
I would work out the pieces of a recurring project, connect them into a process and automate what can be repeated. That could give an internal team, or a team producing websites for clients, a repeatable way to move from research to a working site, with human review built in.

Webinar partner prospecting

Alternative Payments · 2026

Give your team a repeatable way to find the right companies and start the right conversations. I connected research, qualification and outreach so growing a partner program did not depend on doing every step by hand.

10×

my personal prospecting output, built up in less than 60 days.

The interactive demo needs JavaScript. The write-up below covers how it works. Loading the demo…

How it works
Replaced
Finding potential companies without Apollo or Clay, researching their business, identifying the right people and deciding whether they fit. Done thoroughly by hand, that took roughly 1–1.5 hours per company. As the workload grew, keeping that standard became harder.
Built
AI helped source 1,400+ MSP channel vendors from multiple sources. I removed companies that were too large, too small or too close to competitors. Preliminary AI research then produced a Fit Score; companies in the Goldilocks zone moved into Apollo for company and contact enrichment, including LinkedIn profiles. I reviewed contacts for their current company, role and suitability before approving outreach. Lemlist then routed them through campaigns based on factors such as whether we were already connected on LinkedIn.
Result
In less than 60 days, I built a system that helped me review roughly 1,400 vendors, score their fit, choose whom to contact and run an automated outreach process. It grew my personal output to 10× its starting level. The next step would have been to keep increasing capacity while checking that the workload and quality stayed sustainable.
On your team
I would start by reimagining the system for your sales team. The same foundation could support research and outreach across sales, marketing or operations, with the rules and approval steps adapted to each team.

Attribution database and dashboard

Alternative Payments · 2024–2026

Give leadership answers without another spreadsheet chase. I brought data from separate platforms into one reporting system, turning scattered webinar data into a dashboard the team could use to make decisions.

3 systems

connected through about 35 Airtable tables, with a dashboard refreshed daily.

The interactive demo needs JavaScript. The write-up below covers how it works. Loading the demo…

How it works
Replaced
3 systems that don’t share data (Zoom, HubSpot and Salesforce), and no shared view of what the program produced.
Built
An Airtable base of about 35 tables bringing Zoom, HubSpot and Salesforce together. Formula fields handled the calculations, which I audited against the CRM, and a custom dashboard refreshed daily.
Result
Systems that had been separate became a shared view of the webinar program. Leadership could answer important questions about how it was running, with the underlying data available to check.
On your team
I could bring data from your siloed platforms into a custom data warehouse and build reporting around the questions leadership needs to answer, including a few they have not thought to ask yet.

Hush

An automated email triage tool

Help your team find the messages that need a person. I built AI triage that sorts incoming mail and makes its reasoning visible, so routine messages take less attention and important ones are easier to spot.

1 in 9 emails

flagged for my attention, at the current volume of about 60 a day.

The interactive demo needs JavaScript. The write-up below covers how it works. Loading the demo…

How it works
Replaced
A stream of promotions, unsolicited pitches and messages I never signed up for. I was reading them just to decide whether they deserved my time.
Built
A triage system that reviews the sender, subject and body, along with sender-authentication checks such as SPF, DKIM and DMARC. It answers a fixed set of scoring questions, combines the answers into a final score and marks messages above the threshold as important. Time-sensitive messages are surfaced right away; the rest get labels based on rules I continue to refine.
Result
About 60 emails arrive each day, but roughly 1 in 9 is flagged as needing my attention. I can start with that smaller set instead of sorting through everything. I am still tuning the rules to filter more aggressively. In the default mode, Hush adds labels without archiving or deleting mail.
On your team
I could build triage for shared support, sales, service or marketing inboxes. The same pattern could work beyond email: detect low inventory, have an AI agent prepare a draft order, then leave it for a warehouse manager to review and submit. The useful foundation is a system that evaluates incoming information, decides what needs attention and prepares the next step for a person.

Harvest

Turn scattered web pages into source material your team can use. I built Harvest over a weekend to collect a page, a section or a whole site and export readable text for research, purchasing or preparing for a sales call.

No per-page fees

for the research I moved to a custom tool, built over a weekend.

The interactive demo needs JavaScript. The write-up below covers how it works. Loading the demo…

How it works
Replaced
A SaaS tool with credit rules that were hard to keep track of. The service I used charged $5 per 1,000 credits, and a heavy month of collecting pages and whole sites for AI research could use 10,000–20,000 credits. I needed the tool often enough to build my own.
Built
It started as a way to keep my D&D books available without an internet connection. It grew into a web app that takes a URL and reads one page or gathers a site through its sitemap and linked pages. Harvest keeps the main content, leaves out surrounding menus, footers and advertising, and puts the text into Markdown files. More output formats are planned as I need them.
Result
I stopped buying SaaS credits for the work I moved to Harvest. A weekend of building gave me a tool I could keep using and adapt to my own needs, without buying more credits every time I had more pages to collect.
On your team
I could bring your platform’s documentation into the material your AI tools can use. More broadly, I would look for SaaS subscriptions that have become expensive or awkward for the team, identify tools we could realistically replace, and build a pilot around how people want to work. Once it proves useful, we could roll it out and retire the subscription it replaces.

Hosted MCP Servers

Give your team’s AI tools a shared connection to the systems they work with. I build custom connectors that run in one place and can be used across approved devices, with access kept behind a private network.

The interactive demo needs JavaScript. The write-up below covers how it works. Loading the demo…

How it works
Replaced
I was using several AI apps on the same Mac, and many had their own local process for the same connector. Those duplicate MCP processes used resources I wanted for other work. I wanted shared connections that the apps could use without each running another local copy.
Built
I mapped the tools I wanted my AI platforms to connect to, separating provider-hosted connections from those running on my Mac. I moved the local connectors to shared servers in Railway. All of those hosted servers sit behind a private VPN, with access limited to approved devices. That reduces their exposure to the public internet. I also added API capabilities that the standard vendor connectors were missing.
Result
The connectors now run off my laptop. I notice much less CPU and memory in use, which leaves more room for the work I want to do and the other tools I want to run.
On your team
I could give your team’s AI tools shared connections to the systems people already use, with access managed in one place. People could work across approved apps and devices without maintaining duplicate local servers on every computer.

Home Assistant + Parental Controls

Put the controls your team needs in one place. I brought separate household systems into a simple dashboard with buttons that do the work. The same approach can turn a collection of business tools into a control panel built around the people using it.

1 screen

for everyday parental controls, with access reserved for the parents.

The interactive demo needs JavaScript. The write-up below covers how it works. Loading the demo…

How it works
Replaced
Several apps and control screens that made a simple household task confusing. My wife wanted to manage the kids’ devices without having to learn how all the systems worked.
Built
One Home Assistant dashboard for status, schedules and quick changes. I built the missing connection to our parental-control platform and brought in our home network controls, covering Wi-Fi devices as well as devices with cellular data. The children can use Home Assistant too, but permissions reserve these controls for the parents.
Result
My wife can manage the parental controls on her own, from one familiar screen. She is delighted. The kids, now that the controls work properly, are slightly less enthusiastic. Mwahaha.
On your team
I could bring information and actions from separate tools into one custom workspace. A reporting dashboard helps people see what is happening; a control panel lets them do something about it. I would build around your team’s daily work, with the right actions available to the right people.

Mosaic Wadsworth map and poster

Volunteer · 2026

Bring an unusual idea into focus, then make it useful. I learned the mapping tools, worked with AI and gave church leadership clear design options to turn a simple request for a map into a finished campaign poster.

4′2″ × 7′

printed poster with a custom city map and clear campaign instructions.

The interactive demo needs JavaScript. The write-up below covers how it works. Loading the demo…

How it works
Starting point
Years earlier, I had downloaded a basic map and printed it on enormous paper for a church outreach project. When the pastor asked to do it again for a much larger church, I saw room to make something more useful.
Built
I bought map data from NextGIS and worked with AI in QGIS to choose the layers, colors and street labels. A reference grid made streets easier to locate. The map took about 50 hours, including learning the tools. I then built the poster around the church’s visual style, using Claude Design to prepare options leadership could choose from.
Result
A 7-foot-wide poster with a detailed local map, readable instructions and a grid for finding streets. Leadership could choose among prepared options, and we printed the finished piece at the Medina Library Maker Space. The church loved it.
On your team
I could bring that same curiosity and follow-through to an unfamiliar creative project, from a trade-show display to an event experience. I do my best work with a clear goal, room to explore and time to review options together. For a project like this, I would aim to start 30–60 days before it is needed.

24 years of fixing how work gets done

  1. Medina County Career Center

    My first job out of high school. I studied IT at this vocational school, and they were impressed enough with my skills to hire me.

  2. CompSource

    Account manager. I specialized in custom, high-performance computers for CAD/CAM engineering work, and built the templates our sales team used to quote and enter them.

  3. Technology Management Solutions

    Systems engineer at an IT consulting firm, before "managed service provider" was a common term. I was the go-to Juniper NetScreen firewall expert and managed our largest client's Citrix network.

  4. Taylor IT Group

    Owner for 14 years, managing IT for small and midsize organizations. I explain tech so non-technical people understand it.

    Zero breaches, ransomware, or data loss for fully managed clients.

  5. RocketMSP

    Founder and host: 600+ videos, 350,000+ views.

  6. Mosaic Wadsworth

    Tech director. Built and developed an 11-person team from scratch, turning volunteers who weren't audio and video people into a crew that could run it.

  7. Alternative Payments

    First marketing hire. Built the webinar program from nothing: 77 webinars in 19 months, 3,000+ registrations, 1,250+ live attendees, and 145 booked demos. Built the attribution database and partner sourcing pipeline.

  8. Now

    Looking for a full-time role. In the meantime, I'm building projects to keep learning and growing my skills with AI.

Read or print the full resume

Talk about a role

Are you interested in discussing how I can fit into your team? Let’s chat.

Resume LinkedIn

I reply to this address.

A few sentences is plenty: the role, the team, or what you have in mind.