# AI workability — Eastern Migratory Caribou Photographic Survey Interpretation

*Buyer: Department of the Environment (ECCC ) · Source: apc · Closes: 14 Sep 2026 00:00 UTC*

**AI workability: 64%** (majority AI-deliverable) · notice-only read · bid call: **MAYBE**

- Effort: 600 h total · 384 h AI · 216 h human (low confidence)

## Verdict

AI-assisted majority: AI can handle most imagery interpretation and reporting, but human wildlife-expert review is required for final counts.

## Bid call

**MAYBE** — The core work is AI-workable and the shop has relevant data/GIS capabilities, but the missing Statement of Work, unknown evaluation criteria, and need for wildlife-expert QA introduce significant uncertainty. Proceed only if full documents can be obtained and a qualified biologist partner is secured.

## Deliverables

| Deliverable | Mode | AI share | Hours |
| --- | --- | --- | --- |
| Orthoimagery preprocessing and tiling | ai_agent | 90% | 80 |
| Caribou detection and classification | ai_assisted | 75% | 250 |
| Demographic analysis and population estimate | ai_assisted | 70% | 100 |
| Report and documentation | ai_agent | 85% | 60 |
| QA/validation and expert sign-off | human_only | 0% | 70 |
| Project management and client communication | human_only | 10% | 40 |

## Risks

- **[high]** Incomplete Statement of Work / Annex A may hide mandatory methodology or QA requirements. — *Request full solicitation documents and clarify accuracy thresholds, data formats, and deliverable expectations before bidding.*
- **[high]** AI caribou detection accuracy may not meet scientific standards for calves and juveniles. — *Use a human wildlife expert for QA, build a labelled validation set, and run iterative model refinement.*
- **[medium]** Tight timeline from publication to closing limits model development and validation. — *Use pre-trained object detection models and plan an active-learning workflow with expert labels.*
- **[medium]** Imagery volume, resolution, and access method are unknown. — *Clarify data specifications and test on a sample before committing to a processing pipeline.*
- **[medium]** Small shop lacks large reference contracts and may be seen as less credible for scientific work. — *Emphasize AI-agent-first methodology, propose a qualified wildlife biologist subcontractor, and provide relevant data/AI project examples.*

## Human roles

- **Wildlife biologist / remote sensing expert** (32%): Required to validate AI detections, interpret ambiguous imagery, and approve demographic estimates.
- **Project manager / client liaison** (18%): Manages contract, client communication, and approval workflows.
- **Data engineer / AI oversight** (17%): Sets up pipelines, monitors model performance, and handles data integration issues.
- **QA reviewer / validator** (19%): Performs independent checks on AI counts and ensures accuracy against survey objectives.
- **Report approver / scientific authority** (14%): Takes professional responsibility for the final interpretation and population estimates.

## Why not more AI

- AI object detection may struggle with calves and juveniles in dense groups, shadows, or poor image quality.
- No existing labelled caribou dataset is guaranteed; model training and validation require expert-labelled imagery.
- Final population estimates require statistical methodology and scientific defensibility that AI cannot independently certify.
- AI cannot provide professional sign-off or take legal/financial accountability for the survey results.
- Orthographic imagery interpretation may require domain knowledge of caribou behaviour and habitat that AI lacks.

## Requirements & compliance

| Ref | Requirement | Status |
| --- | --- | --- |
| Statement of Work, Annex A | Provide accurate and detailed interpretation of digital orthographic imagery from a June/July 2026 eastern migratory caribou photographic survey in northern Ontario, for the purpose of assessing the total number of adult caribou, calves and juveniles, and other demographic information. | unclear |

## Evaluation criteria

- **Not specified in notice** (Unknown): No evaluation criteria disclosed; likely technical approach and price, but confirmation required.

## Key dates

- 2026-08-19T00:00:00+00:00 — Publication
- 2026-09-14T14:00:00-04:00 — Closing
- 2026-06/2026-07 — Survey period

## Gaps in source documents

- Full Statement of Work / Annex A not provided.
- No evaluation criteria, contract value, or duration disclosed.
- No information on imagery volume, resolution, file formats, or data access method.
- No security clearance, language, or certification requirements stated.
- No details on required report format or delivery milestones beyond the closing date.

## Assumptions

- Imagery is digital and can be accessed remotely; no on-site work is required.
- No security clearance is required for this contract.
- The deliverable is a report and count dataset, not a certified legal or engineering document.
- AI models can be trained or fine-tuned using a subset of expert-labelled imagery.
- Total effort estimate is rough due to missing SOW and data volume details.

---

*Generated by GovContract from public tender documents. Human review required before acting on bid decisions.*