How We Built an AI Workflow to Review Developer and Designer Progress

Introduction

As product teams grow, progress tracking gets noisy. Developers post updates in chat, code ships in GitHub, designers iterate in Figma, and managers spend hours stitching everything together into one status view. We built an AI workflow to automate this review process end-to-end, so daily progress reporting is faster, more accurate, and backed by actual work evidence.

This system ingests developer updates, checks code activity from GitHub, checks design activity from Figma, and generates a structured AI summary that is delivered directly to Slack.

The Core Problem We Wanted to Solve

Traditional progress reporting depends heavily on self-reporting. That often creates gaps:

• Status updates are delayed or incomplete.
• Reported progress is hard to validate against actual commits or design changes.
• Engineering and design outputs are reviewed in separate tools without a unified score.
• Leadership receives too much raw data and too little decision-ready insight.

We wanted a system that could answer one practical question every day: what work was actually completed, by whom, and with what confidence?

Workflow Architecture Overview

We implemented the automation as a workflow pipeline that loops through team members, fetches evidence, runs AI evaluation, and stores/report results in a repeatable structure.

What this flow does: filters relevant team members, fetches historical messages, prepares structured context, evaluates each person with AI, parses machine-readable output, and saves the final evaluation for reporting.

How GitHub and Figma Reviews Are Computed

The AI does not score work from chat text alone. It combines communication with evidence from execution tools:

• GitHub Evidence: PRs, commits, changed files, activity recency, and alignment with shared status updates.
• Figma Evidence: design file updates, component-level iterations, revisions, and designer update consistency.
• Context Layer: role expectations, active sprint priorities, known blockers, and previous-day carry-over tasks.
• Compensation Context: developer experience level and salary band, so effort is evaluated against expected output for that role and cost.

This blended input allows the AI to estimate practical delivery quality instead of only counting raw activity.

Our Evaluation Framework

For each team member, the AI generates:

• Progress Score (0-10): how much meaningful work appears completed.
• Evidence Summary: specific accomplishments validated by code/design activity.
• Risk Notes: inconsistencies, unverifiable claims, missing artifacts, or quality concerns.
• Communication Quality: clarity, consistency, and reliability of updates through the day.
• Value-for-Cost Score: whether delivered output was worth the developer cost, adjusted by experience and role expectations.

Scores are always accompanied by narrative rationale, so managers can quickly understand why a score was assigned.

Was the Work Worth It? (Experience + Salary Lens)

We added a practical efficiency layer to answer a management-level question: did the delivered work justify the spend for that day or sprint?

The AI compares validated output (code quality, completion depth, complexity handled, and consistency) against expected productivity for a developer's experience bracket and compensation range. A senior engineer with a higher salary is expected to produce higher-impact outcomes than an entry-level contributor for the same period.

This gives teams a grounded way to assess ROI per contributor without relying on guesswork, while still keeping final judgment with engineering leadership.

Automated Slack Report Output

After processing all users, the workflow publishes a concise daily report in Slack. Each entry includes the score, validated delivery highlights, and risk signals, making it easy for leadership to take action immediately.

This transforms reporting from manual status collection into a reliable operational dashboard delivered in the channel the team already uses.

Business Impact We Observed

• Faster daily reviews with less manual follow-up.
• Higher trust in progress reporting because updates are evidence-backed.
• Early visibility into delivery risk and communication gaps.
• Unified visibility across engineering and design instead of siloed tracking.
• More focused one-on-one and stand-up conversations.

Most importantly, decision-makers now get a consistent signal every day, rather than fragmented snapshots across tools.

Guardrails and Responsible Use

We treat this system as a decision-support assistant, not an automatic performance verdict. Final judgment remains with team leads. We also built safeguards around data scope, prompt design, and review transparency to keep the process fair and explainable.

AI should amplify managerial clarity, not replace human context.

Final Thoughts

If your team uses GitHub and Figma, you already have most of the data required to build an intelligent progress review layer. The real unlock is combining that evidence with structured AI evaluation and delivering it where teams work every day.

At Craftnotion, we design and implement practical AI workflows like this for real operational use cases, from delivery visibility to automation-driven reporting.

🚀 Want a similar AI workflow for your team? Get in touch and we can help you design, build, and deploy it.