Services
Six ways this
actually helps.
Every offer below traces to a system that is running somewhere right now. They share one engine — specified adversarially, built by verified autonomous workers, gated by a human where being wrong would cost you something.
An agentic system is software that completes work rather than answering questions about it — it plans, calls the tools it needs, checks its own results, and escalates when it is not sure.
This is the core offer. Most businesses that have tried AI have tried a chat assistant: useful for drafting, useless for finishing. The difference is architecture — a system that owns a job end to end needs tool access, memory of what it already did, a way to be wrong safely, and a human gate at the points where being wrong is expensive.
Who it's for
Companies that bought a copilot, saw the demo work, and still have the same manual process running underneath it.
What's delivered
Multi-agent pipelines, tool and MCP integration, autonomous workers behind verification gates, and human-approval ladders sized to how costly a mistake would be.
AI visibility engineering is the practice of making a business legible to the systems that now answer questions about it — so that when someone asks an assistant for a recommendation, your business is in the answer with correct facts attached.
Search stopped being ten blue links and became a paragraph with three names in it. Being absent from that paragraph is a different problem from ranking poorly, and it is fixed differently: structured data an agent can parse, content written to be quoted rather than skimmed, crawler access that does not accidentally block the crawlers that matter, and a review profile that holds up when it gets cited. Where a vertical directory exists for your industry, it anchors the whole thing — visibility without a source of truth underneath it is just louder guessing.
Who it's for
Local and vertical businesses watching discovery move from search results to AI answers.
What's delivered
AI-crawler and citation readiness, llms.txt and schema markup, answer-shaped pages published on a continuous content pipeline, weekly Google Business Profile posts, and review replies — all shipped on your one-click approval. This site is built to the same checklist and audits itself on every change.
Field operations software is the layer between a phone call and a signed job — capture, qualification, estimate, approval, schedule.
For service businesses, the lead is usually not lost to a competitor with better marketing. It is lost to a voicemail nobody returned and an estimate that took four days to write. The fix is to make capture instant and drafting automatic, while leaving every number that goes out the door under the owner’s thumb.
Who it's for
Home and field-service businesses where the owner is also the estimator, the dispatcher, and the person driving the truck.
What's delivered
Capture surfaces, agent-drafted estimates with an explicit human approval step, and per-tenant deployments so each business owns its own data.
Communications automation is what keeps a business reachable when no one can pick up the phone — the text that goes out on a missed call, the review request after the job, the check-in before the appointment.
The mechanics are simple and the compliance is not. Consent has to be recorded, opt-outs have to be honored everywhere and immediately, and messages have to respect quiet hours. Building that ledger correctly once is the entire difference between a channel that works and one that gets a number shut off.
Who it's for
Businesses that lose work to missed calls, and anyone who wants texting done without inheriting a compliance problem.
What's delivered
Missed-call text-back, review requests, pre-service check-ins, and AI answering — one number, one consent ledger, opt-out and quiet-hours enforcement built in.
Support continuity is the part of a system that handles the moment it breaks — capturing what actually happened while the person is still in front of it, instead of three days later in a vague email.
This offer exists because a client asked a fair question: what happens when something goes wrong? Most incident reports arrive as "it stopped working," which starts a week of guessing. A report that arrives with the state of the system attached starts a fix instead.
Who it's for
Anyone running software they depend on, especially where the people hitting the bug are not the people who can describe it.
What's delivered
Structured incident capture at the moment of failure, diagnostics-rich reports, and a fast loop from report to fix.
A vertical AI directory is a structured, agent-readable index of one industry — built so an assistant can query it directly and answer with verified facts instead of guessing from scraped pages.
Generic directories were built for people browsing. Agents need something else: real structure, current data, and an endpoint they can call. But the directory is only the first layer — once an industry's data is trustworthy, it can be turned back on the operator as recurring intelligence, and eventually run the operations it describes. That progression is the product.
Who it's for
Industries where AI assistants are already making recommendations from data nobody is maintaining — and operators inside those industries who are flying on last month's numbers.
What's delivered
MCP endpoints, structured answers, and operator claim flows on the Tenet platform, plus recurring intelligence built from that data. Live for Twin Cities dining through Misen, with dental in build through Tendant.
How an engagement works
Conversation
We talk about your problem. I ask questions and listen. No pitch, no agenda.
Honest assessment
I tell you what I think is possible, what it would take, and whether I'm the right person for it. If I'm not, I'll say so — including when the honest answer is that you don't need AI for this.
Scoped slice
The work is cut into a slice and specified to binary success criteria before code is written. You see the board: what each task must make true, and what "done" means, in advance rather than in hindsight.
Long-term support
These systems are meant to run for years, and I stay involved to make sure they do — including structured incident capture when something breaks.
Common questions
What kinds of work are a good fit for AI automation?
Work that is high-volume, repetitive, follows a pattern, requires specific domain knowledge, and produces a structured output. RFP responses and general-ledger coding are textbook examples; judgment-heavy work such as strategy, naming, and design is not.
Do I have to replace the software my team already uses?
No. The best systems are invisible: they show up in the inbox, the phone, or the tool your team already opens, rather than demanding a behavior change. Adoption failure is the most common way good software dies.
How does an engagement start?
With a conversation about the problem and an honest assessment of whether it is worth solving this way. If it is, the work is scoped into a slice with binary success criteria written down before any code exists, so you can see what "done" means in advance.
What happens after the system ships?
The systems are built to run for years, so support is part of the offer rather than an upsell: structured incident capture when something breaks, and ongoing changes as the business changes.
A note on what I won't do
I won't take on a project I don't believe in. I won't build something just because someone is willing to pay for it. And I won't oversell what AI can do — the technology is genuinely powerful, but it isn't magic, and clients who come in with realistic expectations get far better outcomes than those who've been promised the moon.
The businesses I work best with tend to have a specific problem they want solved, an openness to thinking differently about how it gets solved, and a realistic sense that good systems take some investment to build and some time to prove themselves.
If one of those sounds like you, let's talk.
Every engagement starts the same way: a conversation where I try to understand your problem well enough to be honest about whether I can help.
Get in touch