Start a conversationContact

Systems built around real businesses.

Three client relationships, four systems, and one capability engagement. Each case gives the client type, the operating problem, the system, what people can now do, and where that work stands today.

These are systems built for a particular operation, not products on a shelf. What you can obtain from Bedrock is set out under What We Build.

What you can obtain

Regional homebuilder and land developer

An operating system built around the business

Land, construction, sales and finance ran on definitions that disagreed, so the same question got a different answer in each department.

One operating state every tool reads from, a console where the recurring work runs, and a queue that turns each blocker into a decision with a named owner.

A controller can see whether a phase is ready to close and what is holding it. An operator clears a blocker by making the one decision it needs, and the record of that decision stays with the work.

In production in the client’s environment, operator-run with handoff to the finance team in progress. Verified against the repository and an outside technical review in July 2026.

Read the case
Phase readiness: every community with a readiness verdict and the four cost components as chips, four of them reportable, one partial because its indirect cost pool is out of date, one blocked — and on a wide screen the owner of each gap and the next action beside it.
Phase readiness across the portfolio: which communities can report, and which parts the rest are still missingClient systemApplication shown with synthetic data

One relationship, two systems

Two systems built for a single commercial real estate operator, in a relationship that has since ended. They are not two versions of one idea: one runs the buildings the operator already holds, the other decides which ground to buy next.

The operations board: open work orders set out in columns by the state they are in, each card carrying its number and what is wrong, with a due date once one is set — the whole portfolio across on a wide screen, the newest requests at the front on a phone.
The operations board: every open job across the portfolio, in the states the work moves throughClient systemApplication shown with synthetic data

Commercial real estate operator

Property operations

Maintenance ran on a board beside the accounting system, and a repair counted as done when the technician said so.

One application for the operations desk, the technician in the field, the resident and the outside vendor, on one record per request that all of them work from — and a repair finished only when the resident says it is.

While it ran, a coordinator could see every open job across the portfolio, assign it against the vendor’s insurance, and follow it to resolved — and could close it only once the resident had answered.

Deployed during a completed engagement, May to September 2026. Bedrock does not operate it now.

Read the case
The acquisition workspace: a saved buy box down the left, a county of parcels on satellite imagery coloured by zoning verdict in the middle, and a ranked candidate table on the right with score, parcel number, address, zone, storage use and acreage.
The acquisition workspace: a county screened against a written buy box, ranked for the analystClient systemApplication shown with synthetic data

Commercial real estate operator

Acquisition intelligence

Every parcel triggered a fresh hunt for the governing municipality and its ordinance, and nothing learned on one parcel carried to the next.

Jurisdiction research kept as a system of record, candidates screened in bulk against a written buy box, and an analyst adjudicating in batches.

While it ran, an analyst could screen a batch of candidates against the buy box, see exactly where the ordinance stopped, and decide the cases the rules could not, with the next parcel in that county inheriting the answer.

Deployed and run at scale during a completed engagement. Bedrock does not operate it now and makes no claim about its use since.

Read the case
The leadership view: one community intelligence score out of a hundred, beside the confidence in the evidence behind it, the coverage, how fresh the corpus is and how large; then four figures, each carrying a confidence badge with its sample size, the number of observations behind it and a drill into them; then what changed this week, each delta with its own evidence count.
The week’s read from fifteen debriefs: what changed, and how much evidence sits behind each lineIn developmentApplication shown with synthetic data

Large-scale mixed-use community

Community intelligence

What a community learns about its buyers sits outside the operating picture leadership has, and a handful of remarks gets reported as a trend.

Structured capture on a phone, shared definitions, and a rule that holds a claim back until more than one source has seen it.

A host can record a tour debrief in a fixed shape, and every figure in the weekly read carries the count of observations behind it and opens to the records it was counted from. Exercised against synthetic data only; no host uses it on real tours today.

In development. One center is live in the production console and deliberately blocked with no data loaded. The capture and evidence tools are built and exercised against a synthetic corpus only; no host uses them today.

An initiative inside the homebuilder relationship

Read the case

Field note · not a case study

Industrial engineering and manufacturing company

An AI capability engagement

The company held licences for AI tools and no shared standard for using them on its own work.

Four weeks at a fixed fee, quoted in the first conversation: shared standards, a working session per function, a redesign week and a build week, ending in a readout where the company’s own people demonstrate what they built.

What its people can do at the end is for the readout to show. Nothing is claimed before it.

Kickoff held September 2026. A four-week capability engagement at a fixed fee, so nothing has been deployed and nothing has been measured.

How an engagement runs
The four-week arcKickoff held · readout not yet held
  1. Week 1By FridayShared standards · sessions per function begin
  2. Week 2By FridaySessions continue · candidates listed per function
  3. Week 3By FridayRedesign week · one workflow defined end to end
  4. Week 4By FridayBuild week · readout: build, buy, wait or skip on each candidate

If a number appears in the readout, it was measured during the engagement — or it does not appear.

What the company keeps
  • Executive briefing
  • Shared workspace, installed and furnished
  • Prompt library, each entry owned
  • Prompting standards, one page
  • Workflow shortlists, ranked per team
  • Roadmap
  • Recording and materials library
  • Office-hours answers
  • Build, buy, wait or skip calls
  • Final readout

Kickoff held September 2026. A four-week capability engagement at a fixed fee, so nothing has been deployed and nothing has been measured.

Illustrative · the engagement’s shape, from the proposal

None of these engagements carries usage instrumentation, so there are no hours, adoption or dollar figures to report, and none has been estimated in their place.

Bring us an operating problem, a new opportunity, or an AI mandate that needs turning into work people can do.

Start a conversationadam@bedrockos.co