Ideas

    What we'd build in your field.

    These are short cases, one field at a time. Where we've built the thing, the label says built. Where we haven't, it says so, and we describe the build we would run.

    Client names stay out. We work inside our clients' companies, and their privacy is part of the job. The page opens with a point of view from the team.

    Professional servicesPoint of view

    The hourly consulting model is ending. Here is what an AI-native engagement looks like instead.

    Sam Westfall, September 16, 2026. First published on LinkedIn

    Deloitte told its own consultants in May that the hourly consulting model is finished.

    A partner in the US public sector practice put a chart on screen at an internal town hall showing hours-based consulting shrinking to a sliver of the professional services market by 2035, with AI agents taking the rest, and when the Wall Street Journal reviewed the recording one consultant in the room summed it up for them: "They heavily implied our model is toast."

    The chart shows the end of the model, so it's worth saying what the model looked like from the client's side while it was still working.

    A consultancy has to feed a very expensive person at $350 an hour billed by the man-day, which means the engagement is built to keep that person fed, so scope grows, the team grows, the second phase gets sold in month three, and by the end you've spent $500K on a deck about unlocking your potential.

    Anyone two levels below the executive who signed will tell you the consultant was selling cover for a decision nobody wanted to own.

    There is a version of the hourly model that earned its money, which is the specialist at $10K an hour who unblocks the system that's been strangling your operations for two years, and I've paid for that trade without complaint.

    Generic advice was where the hour was always a trap, and now a model hands you the framework in thirty seconds, so if your strategy needs a $350-an-hour outsider to tell you what it is, the honest move is to change jobs.

    Deloitte's own 2026 enterprise survey explains why the chart looks the way it does, since 84% of companies haven't redesigned a single job around AI and only 21% have a mature way to govern agents, while nearly three quarters plan to deploy them within two years.

    Closing that gap is engineering work rather than advisory work, and engineering stopped billing by the hour the moment 98% of the code we ship stopped being handwritten.

    So the question I'd put to any vendor claiming to be AI-native is how many people they've let go and why they're hiring more, because an integration for us is one senior engineer, agents across the delivery cycle, and a fractional architect, about 1.5 FTE where a five-person team used to sit, at $5-$15K a month with proof of value in month one.

    Every engagement we run has fewer people on it now than when it started, because the harness takes more of the load as the use cases get done, and we've never grown faster.

    McKinsey says a third of its fees are already tied to outcomes, Deloitte just told its people the other two thirds are heading the same way, and the firms still standing in 2035 will be the ones that were shipping while the rest were preparing slide decks.

    Property managementBuilt

    Owner statements and rent follow-up that run without a person chasing them.

    The problem, in the owner's words

    Owner statements get put together by hand every month. Rent follow-up happens when somebody remembers. Maintenance requests come in by email and get sorted one at a time, and the same tenant questions get answered over and over.

    What we built

    • Rent follow-up and late payment sequences that run on their own.
    • Owner statements and reports that generate from the books.
    • Maintenance requests taken in and sorted before a person touches them.
    • Tenant messages handled by AI, with handoff rules the team controls.

    What the owner holds at the end

    You hold the system and the data in it. Your team works the exceptions, and the routine work no longer depends on anyone's memory.

    ConstructionThe build we would run

    One job record for the bid, every change order and every photo.

    The problem, in the owner's words

    The bid went out from one tool, the change orders live in email, and the job photos are on the super's phone. When a customer disputes a bill, you dig through all of it to rebuild what happened.

    What we would build

    • A job record that holds the bid, each change order and each photo in one place.
    • An agent that drafts the change order from the field note and sends it to the project manager to approve.
    • A job cost report that builds itself from your own numbers.
    • All of it on a database you own, so the history stays with the company.

    What the owner holds at the end

    You'd hold your company's job history in one database you own, and the tools that read from it.

    The tradesBuilt

    Lead follow-up that doesn't wait for the owner to get off a roof.

    The problem, in the owner's words

    Leads come in from the website, the phone and a few lead sites. Each one gets a single call. Then the owner is back in the field and the lead goes cold. Old quotes never get a second look.

    What we built

    • Lead capture from every source into one pipeline.
    • An AI reply by text and voice as soon as a lead comes in.
    • Follow-up that keeps going after the first attempt.
    • Booking straight onto the calendar, and a review request after the job.

    What the owner holds at the end

    You hold one pipeline where there used to be a handful of inboxes. Follow-up runs by system, so it no longer depends on who remembered.

    Private equity deal sourcingBuilt

    A target list that doesn't start from a stale database.

    The problem, in the owner's words

    Every firm says it has its own deal flow. Few have a system that produces it. Analysts spend weeks building target lists from stale databases, and the outreach dies in unanswered email.

    What we built

    • Business discovery across a whole market, with every target profiled for ownership, size signals and growth signals.
    • Outreach managed by AI across text, email, voice and iMessage.
    • AI holds the first conversation, and a person approves every message that matters.
    • Every reply and every meeting tracked back to the campaign that produced it.

    What the owner holds at the end

    The firm holds its own sourcing system and its own data. The pipeline stays when an analyst leaves.

    Commercial real estateBuilt

    A pro forma from the deal terms, without rebuilding the model.

    The problem, in the owner's words

    Every deal starts with a pro forma. Every pro forma starts with an analyst rebuilding the same model by hand, from comps and rent assumptions to construction costs and financing terms. The deal can be gone before underwriting has an answer.

    What we built

    • Deal terms go in and a full pro forma comes out.
    • Market data and comps pulled in automatically.
    • The firm's own assumptions entered once and applied the same way to every deal.
    • Sensitivity cases generated beside the base case, laid out for lenders and investment committees.

    What the owner holds at the end

    The developer holds a model that carries its own assumptions. Every deal gets underwritten the same way, no matter who runs it.

    HealthcareBuilt

    A CRM built around the patient journey.

    The problem, in the owner's words

    A behavioral health operator runs on referrals, and referrals die in handoffs. One comes in by fax. Another lands in a spreadsheet. The records system was built for charting, so nobody can say which referral sources lead to admissions.

    What we built

    • Referral source tracking that shows which relationships produce admissions.
    • An intake pipeline with insurance verification status at every step.
    • Automatic follow-up on stalled admissions, so a referral doesn't quietly expire.
    • Built to the operator's workflow and owned outright.

    What the owner holds at the end

    The operator owns the system, the data and the playbook. One record runs from referral to admission.

    If your field isn't on this page, tell us the job. Talk to the team and we'll tell you if we've built something like it.

    Start here

    Tell us what should
    run itself.

    We'll tell you plainly whether a build fits, and what you'd own at the end.

    Talk to the TeamSee what we build →