Evaluate a decision
Compare a candidate with your current method using agreed examples, quality checks and complete review effort. A recommendation to wait can be a useful result.
Offering · Primary engagement
Is AI worth it
for this task?
Before committing to a tool, hardware or a build, compare one AI candidate with the way your team works today. CAD 2,500 buys an agreed test brief, a comparison record and a written recommendation, including a recommendation to keep your current method. Start with a free call to decide whether that comparison is useful. Local-first delivery is our specialty; hosted and frontier options are considered where your data requirements allow them.
Your assessment packet
From a question
to a decision.
Before booking, we confirm the question, permitted inputs, boundaries, reviewer and fee in writing. The detailed test brief is part of paid delivery. We confirm a start date when the agreed inputs and access are ready; the one-week schedule begins on that date.
- 1. Test brief. The task, current method, candidate, examples and criteria.
- 2. Comparison record. Results, errors, review effort and operating-cost assumptions.
- 3. Written recommendation. What the evidence supports, unresolved questions and the next decision. If the comparison is inconclusive, the report explains why and identifies the smallest useful next check.
You keep the packet whether you proceed, stop or retain your current method. Testing beyond the agreed scope and any implementation require a separate agreement.
Who it's for · 01
Who this
is for.
For a business owner, operations team or IT partner with a recurring task to improve. The assessment needs an accountable owner, usable examples and a result you can judge.
- 01
A business task or AI opportunity you want help evaluating, with representative inputs, a named owner and a useful result to test.
- 02
A clear reason to control where AI processing happens, with data-handling requirements that can be specified.
- 03
Existing systems that need a connector or a workflow-specific output, such as structured records, evidence-backed search or reviewable drafts.
- 04
An internal technical owner or managed IT partner who can help operate and maintain the delivered system.
Before we start · 02
Before we
start.
Before booking, we agree the task, the examples we can use and how to judge the result. These three foundations give the assessment a useful starting point.
- 01
Examples we can use. Documents, records or other inputs you have permission to share. We agree whether real examples, samples or synthetic data are suitable for the question.
- 02
Someone who knows the work. A person who can explain the current process and help judge whether the results are useful.
- 03
An agreed question and scope. We define what the one-week assessment will test. Building a production system and providing ongoing support are separate scopes.
The comparison should inform a decision worth testing. If an existing tool or a simple trial can settle it adequately, we may recommend that instead. A paid assessment can recommend keeping your current method; an inconclusive result explains what remains unknown. Testing beyond the agreed scope is agreed separately.
Working together
Useful at each
stage of adoption.
A comparison, an integration and an operational handover solve different problems. We agree the work that supports your next decision.
Integrate and hand over
Connect the agreed workflow, document its limits and recovery, and train the people responsible for it. Existing tools and your technical team remain part of the design.
Define continuing duties
Where follow-up is useful, distinguish evaluating a proposed change from maintaining a deployed system. Agree covered work, hours and responsibilities before making a continuing commitment.
Reusable tests and operating notes help your team understand the result. Collaboration with your existing provider can be scoped with clear delivery and support ownership.
Choose the approach
What needs
to change?
Start with the task. The right approach may combine several of these, or use simpler software you already have.
Find the right information
Connect answers to approved documents with sources and access controls. Retrieval helps a model use relevant information; it does not change the model's underlying skills.
Connect tools and complete steps
Integrate a bounded workflow with your existing systems. Agree what it may read or change, what needs human approval, and how to recover when a step fails.
Improve a model on a repeated task
Compare clearer instructions and retrieval with adapting an existing model. A training experiment needs permitted data, independent test examples, a compute budget and an agreed improvement target. Training is proposed only when that comparison justifies it.
Help people operate the workflow
Train staff to use the tools, check the output and handle exceptions. This is training people, distinct from training a model. Handover and any further support are agreed in the scope.
You do not need to buy anything beyond the assessment. Ongoing evaluation or maintenance is a separate decision based on a named responsibility, an agreed budget and evidence that it is useful.
Engagement · 03
A clear scope.
At each step.
Choose the work your decision needs. The free call establishes an initial view of fit. Buy the comparison when evidence is missing; proceed directly to a scoped build proposal when enough evidence already exists. These are options, not required paid steps.
01
Up to 45 minutes
Discovery call
Free
Discuss one task, how it is done and checked today, data constraints, who owns the decision, timing and what would make a change worthwhile. Receive a short email recap of the task, our initial view of fit and the agreed next step. The call does not include feasibility testing or architecture. Where existing evidence is sufficient, we can prepare a build proposal directly. No obligation to buy.
02
1 week · fixed-fee
AI workflow assessment
CAD $2,500
Answer one agreed buying question using permitted examples, a current-method baseline and one candidate. Compare quality, human review effort and operating cost. Receive the agreed test brief, comparison results, a record of errors and review effort, and a written recommendation with limits and next steps. Scope and any specialist review are agreed before booking. The fee is credited in full toward a fit-and-scope assessment or build within 90 days; each credit is applied once. HST/GST is added where applicable.
03
Scoped per project
Implementation, when ready
Separately quoted
When the evidence, acceptance criteria and dependencies are clear, we quote the build, integration, training and handover. A further fit-and-scope assessment is needed only to resolve open implementation questions; it reuses earlier findings. Support is agreed separately. You can also stop or keep your current method.
Prices are in CAD, with HST/GST added where applicable. Credit terms and payment milestones are recorded in the agreement.
Where we buy deployment hardware for you, it is billed at cost plus a 5% handling fee.
Data control · 04
Three layers.
Specified in writing.
For each engagement we document three things separately: your content, the connection records around it, and who administers the system. The design states which controls it provides, which outside dependencies remain and who is responsible for them.
Local hosting is the default and one part of the complete data-control design. Data classification determines whether a hosted option is acceptable.
03
Administration
Who can control the system?
Routing
Keys
Access & administration
01
Your content
What the workflow processes
DocumentsApproved inputs
AI workflowPrompts & processing
ResultsResponses for review
Specify where content is processed, stored and backed up.
01
Your content
The content processed by the workflow. Identify where it is stored, transmitted and used for inference, including connectors, logs and backups. Define permitted flows and controls in the delivery specification.
02
Connection records
Connection identities, timing and usage records may be visible to network or service providers even when content is encrypted. Document which parties can observe these records and how retention is handled.
03
Administration
Identify who operates connection routing, manages keys, grants access and can revoke it. Document third-party dependencies and recovery responsibilities alongside the selected infrastructure.
Deliverables · 05
What a build
can include.
The assessment delivers its findings and recommendation. For a build, the signed scope sets out exactly what you receive. Typically that includes:
The working workflow, configured for your task and tested against the agreed checks.
Connections to your data sources, with the source code handed over.
The AI model set up where the agreed design specifies: on your own hardware by default, or a hosted service where your data requirements allow it.
Output in the format your team already uses.
Operating instructions covering day-to-day use, troubleshooting and recovery.
A training session for the person who will own it.
And what you don't get.
A closed platform you have to keep paying for.
A system you can't inspect or change.
Any cloud service or outside connection you haven't approved.
Discovery call
One task. Up to 45 minutes. An agreed next step.
Request a free discovery call Tell us about your task Up to 45 minutes · free · no obligation
See the contact page to send a short request. We arrange a time by email; sending a message does not book paid work.