AI Decision Support Systems with a Human Decision-Maker
An AI decision support system prepares a case for a person: it gathers the facts, applies your criteria, drafts a recommendation with reasons and logs the final human decision. Remolda builds one in six weeks for $9,800 CAD + HST.
In short
- Price: $9,800 CAD + HST for the six-week AI Pilot Sprint: one decision process supported in production.
- The system assembles the file, checks it against your written criteria, drafts a recommendation with reasons and records what the person decided.
- Built for reviewable decisions: eligibility, triage, credit or risk review, contract review, resource allocation.
- Designed around Canadian rules on automated decisions: Quebec Law 25 s. 12.1 and, for federal institutions, the Directive on Automated Decision-Making.
- Every recommendation shows its sources, so reviewers can check it in minutes.
Your situation
Decision support fits teams that make the same kind of judgement many times a week. Typical cases:
- Files take hours to assemble. The decision is quick; gathering the facts is slow.
- Different reviewers, different outcomes. Similar cases get different answers, and appeals follow.
- Criteria live in people's heads. A senior reviewer retires and the reasoning goes with them.
- Auditors ask why. Reasons for past decisions are hard to reconstruct.
What the system does
| Step | System | Person |
|---|---|---|
| Gather | Pulls records, documents and history into one case view | Adds context the system cannot see |
| Check | Tests the case against written criteria, flags gaps | Confirms or corrects the flags |
| Recommend | Drafts an outcome with reasons and sources | Decides and signs |
| Record | Logs recommendation, decision and differences | Explains overrides in one line |
| Learn | Reports agreement and override patterns | Updates criteria when needed |
Where the input is mostly documents, the gathering step uses intelligent document processing. Scores or forecasts feeding the case come from predictive analytics.
Canadian rules that shape the design
- Quebec Law 25, s. 12.1. A decision about a person based exclusively on automated processing must be disclosed to that person, who can ask for the information used and the principal factors, and submit observations to staff able to review the decision.
- Directive on Automated Decision-Making. Federal institutions complete and publish an Algorithmic Impact Assessment before production; impact is rated from Level I to Level IV.
- Bill C-36. Tabled in June 2026, it proposes transparency duties for automated decision-making about individuals. It is not law yet.
Details for your sector are covered in AI compliance in Canada.
What the AI Pilot Sprint includes
Criteria written down. Your rules and senior reviewers' reasoning turned into a checklist the system and people both use.
One decision process in production. Case view, checks, draft recommendation and decision log.
Calibration and override tracking. Measures that show whether reviewers still judge independently.
Disclosure and records. Notice text and log fields that match the rules that apply to you.
Training and runbook. For reviewers and for the person who owns the criteria.
How long it takes
Six weeks from kickoff. In our experience the usual split is two weeks for criteria and access, two weeks of build and testing on past cases, and two weeks of live use with review. Legal or privacy review of the notice text can add time.
What it costs
The AI Pilot Sprint is $9,800 CAD + HST, fixed, for one decision process. Model and hosting usage is billed by the vendor. If you first need to decide which decisions to support, the AI Opportunity Audit ($2,900) is credited toward the Sprint. See the pricing page.
Why Remolda for decision support
- A person decides. The design keeps a named human accountable.
- Reasons are logged. Every recommendation and override is recorded.
- Canadian rules built in. Law 25, the federal Directive and your AI use policy.
- Fixed price and scope. One process, six weeks.
- English and French. Case views and notices in both languages.
How the work runs
The Sprint covers the Strategy and Implement steps of the Remolda Cycle (Audit → Strategy → Implement → Empower → Evolve).
- Scope call. 30 minutes: which decision, how often, who signs.
- Criteria and past cases. Written rules, a sample of past files.
- Build and back-test. The system runs on past cases; results compared with actual outcomes.
- Go live. Reviewers use it daily with override tracking.
- Review. Time per case and consistency against the baseline, then go / adjust / stop.
Frequently asked questions
What is an AI decision support system?
It is software that prepares a decision for a person. It collects the relevant records, applies the organization's criteria, highlights risks and drafts a recommendation with reasons. A named person makes and signs the decision, and the system logs both.
How much does an AI decision support system cost?
Remolda's six-week AI Pilot Sprint is $9,800 CAD + HST for one decision process in production, including criteria mapping, the build, reviewer training and a runbook. Model and hosting usage is billed by the vendor.
Does Quebec Law 25 apply to AI-supported decisions?
Section 12.1 applies when a decision about a person is based exclusively on automated processing: the person must be informed, can ask for the information used and the main factors, and can submit observations to staff who can review the decision. A human decision-maker changes the analysis; we document the design either way.
What does the federal Directive on Automated Decision-Making require?
For federal institutions, the Directive requires an Algorithmic Impact Assessment, published before an automated decision system goes into production, and sets four impact levels with matching safeguards. It does not bind private companies.
How do you prevent reviewers from rubber-stamping the AI?
Recommendations show evidence and uncertainty, some cases are routed without a recommendation for calibration, and the log tracks how often reviewers change the outcome. These figures are reviewed monthly.
Which decisions are a good fit?
Frequent decisions with written criteria and a record of past outcomes: benefit or program eligibility, intake triage, supplier or credit risk review, contract clause review.
How long does it take?
Six weeks from kickoff. In our experience, writing down the criteria the team actually uses takes longer than the build.
Sources
- Treasury Board of Canada Secretariat — Directive on Automated Decision-Making
- LégisQuébec — Act respecting the protection of personal information in the private sector (P-39.1), s. 12.1
- Parliament of Canada — Bill C-36 (45th Parliament), LEGISinfo
- Office of the Privacy Commissioner of Canada — AI, privacy, and your business
Facts checked:
Related services
Approach phases
Related insights
LLM Integration into Existing CRM and ERP Software in Canada: Architecture, Data Residency and Cost
Mitacs AI Advantage: Ottawa Commits $162M to 10,000 AI Work Placements
AI for Canadian Municipalities: Where It Works in 2026
Talk to an AI transformation consultant
A 30-minute call: you describe the situation, we tell you what to do first and what it would cost.
Book a 30-min call30 minutes. English or French.