Applied-AI Systems

For construction and industrial teams.

Choose one task that costs your team time or money each week. I test it with your data, measure the result, and build a system your team can use and review.

Each workflow has a named owner and a clear acceptance test. Human review stays in place where judgment matters.

Stages and prices

1. Workflow Review

CAD $10,000 · Two weeksTest the data. Define success. Decide: go, revise, or stop.

2. Working Pilot

From CAD $35,000 · Six weeksBuild one workflow. Measure it against the acceptance test.

3. Production and Managed Operation

From CAD $75,000
plus CAD $9,000 per monthDeploy the accepted workflow. Monitor results and improve it each month.

Move forward only when the evidence supports it. A pilot needs separate approval before it enters production.

Start with one workflow

Good starting points repeat often, have a visible cost or delay, and already have a person who owns the result. Examples include:

• extracting submittal requirements from project specifications;
• qualifying RFPs against bid criteria;
• searching private project knowledge;
• flagging contract clauses for human review;
• checking an estimate against past project data;
• turning field records into a management exception report.

Do not start with a general chatbot or a mandate to "use AI." Start with one input, one output, one owner, and one acceptance test.

Who this is for

Proceed when all six conditions are true:

1. A named executive owns the result and can approve process changes.
2. The workflow repeats often enough to measure within 90 days.
3. The current cost, delay, error, or risk is worth at least CAD $250,000 per year.
4. The client can provide representative source data and a process owner.
5. One acceptance test can define a useful result before development starts.
6. Human review stays in place for safety, legal, financial, and contractual decisions.

This is not a fit for discovery-only work, speculative chatbots, 24/7 support, or autonomous high-consequence decisions.

Three stages

Workflow Review

CAD $10,000. Two weeks.

Map the current workflow and failures, test representative data, measure the baseline, define the acceptance test, and set the data, access, human-review, and ownership boundary. Finish with a go, revise, or stop recommendation.

Working Pilot

From CAD $35,000. Six weeks.

Build one workflow on representative data. Deliver its evaluation set, review controls, operator guide, and measured result against the baseline. Keep the pilot out of production unless the client separately approves production access and obligations.

Production and Managed Operation

From CAD $75,000 plus CAD $9,000 per month.

Deploy one accepted workflow into the client's approved environment. The managed operation monitors its acceptance checks, issues a monthly reliability and value report, ships one bounded improvement each month, and holds one operator review.

New workflows, major integrations, migrations, 24/7 support, and material scope changes are separate work. The client owns its production accounts and pays its infrastructure and model providers directly.

Proof before expansion

Each stage ends with a decision. Keep the evidence visible:

• the baseline source;
• the fixed evaluation set;
• the acceptance result;
• known failures and unresolved risks;
• measured delivery time and operating cost;
• a go, revise, or stop recommendation.

If the pilot fails, stop or revise it. Do not move it into production because the demo looks useful. If the managed service exceeds its scope or capacity, reduce the scope, improve the process, or price the added work separately.

Common questions

Why only one workflow?

A narrow boundary makes the result measurable, limits access, and keeps delivery viable for a solo practice. A second workflow is a second scope decision.

Who owns the data and accounts?

The client owns its data and production accounts. Access, retention, security, privacy, and legal duties belong in the signed agreement and approved production plan.

Does the system make the final decision?

Not for safety, legal, financial, or contractual judgment. The system can assemble evidence and flag exceptions. The named human owner remains responsible.

What happens after launch?

The monthly review records operating results, failures, scope, capacity, one improvement, and a keep, change, or stop decision. Claims stay tied to the evidence that supports them.

Bring one workflow

Send the workflow, current tools or data, result needed, and deadline. I will tell you whether it is measurable enough for a review.