Replace legacy tools with agent-native alternatives

You don't replace Excel — you put an agent on top of it. The spreadsheet remains the output format humans are familiar with, but the agent handles the data cleaning, formula writing, and analysis that used to require deep Excel expertise. Claude Code + Google Sheets MCP turns 'build a VLOOKUP across 3 sheets' into 'compare these two data sets and highlight differences.' The LLM's natural language interface is genuinely better than function syntax for 90% of analysis tasks.

Personal knowledge SaaS products are the clearest case of 'LLM does it better through natural language.' Recall charges $15-30/month for save-and-search. An LLM-powered agent + Obsidian gives you save, organize, summarize, cross-reference, generate insights, and weekly digests — for $2-5/month in API costs. The setup is 2-4 hours, but the experience is dramatically better because the LLM understands your knowledge contextually, not just keyword-matching. If you're already using Claude Code, you're 90% of the way there.

Traditional Figma handoff is a lossy process that doesn't scale with agentic coding speed. The Figma MCP + Storybook MCP + Tokens Studio stack creates a machine-readable bridge from design to code. monday.com's production deployment proves it works at scale — their 11-node agent architecture generates code that 'looks like someone who deeply understands the system wrote it.' Start with Figma MCP (free, 5 minutes), add Storybook MCP for component context, and Tokens Studio for bidirectional token sync.

Manual design review should be the LAST gate, not the only one. Automated visual regression catches 80% of drift before a human ever sees the PR. The human reviewer's job shifts from 'catch visual bugs' to 'approve intentional design changes' — a much higher-value activity. Start with Playwright VRT (free, 5 minutes to set up) and add Percy/Chromatic when you need cross-browser or team features.

Don't replace Terraform — augment the human operating it. The agent handles the tedious parts (plan review, drift detection, boilerplate generation) while humans keep control of production changes. Full auto-provisioning is possible but only for non-critical environments.

Replace your most brittle Selenium tests now. The maintenance savings are immediate. Keep Selenium for performance-critical test suites and CI smoke tests where speed matters. The ideal setup is a hybrid: agents for functional coverage, Selenium for speed.

Migrate anything beyond a single-command cron job. The observability and error handling improvements alone are worth it. Keep trivial crons (log rotation, temp cleanup) where they are — they don't need an agent.

Don't fully replace Notion — augment it. Notion is still the best tool for human knowledge work. But agents need structured, queryable memory, and Notion isn't that. Run both: Notion for humans, memory system for agents.

High-effort, high-reward. The agent-native support stack is genuinely better for companies doing 500+ tickets/month. Under that volume, Zendesk is fine. The key risk is HITL design — bad escalation flows destroy customer trust. Get that right and the ROI is 3-5x.

Partial replacement is the right move. Use Langfuse or Helicone for your AI/agent layer. Keep Datadog for traditional infrastructure unless your stack is 90%+ AI workloads. The cost savings on the AI monitoring alone usually justify the migration.

This is the easiest win on the list. Every team should have an agent reviewer as of yesterday. It doesn't replace human review — it makes human review focus on what humans are actually good at. Start today.

Hard migration but worth it for teams under 200 people who are tired of Jira's complexity. The agent integration opportunity with Linear is genuinely better — Linear's API was designed for this. Enterprise teams with heavy Jira customization should stay put.

The agent-native monitoring space is early. Replace PagerDuty if you're drowning in alert fatigue and have engineering capacity to build the triage layer. If compliance is a hard requirement, keep PagerDuty as the outer shell and use agents for intelligent pre-filtering.

Worth it for teams running 50+ Zaps or hitting pricing walls. The migration is real work — budget 2-3 days for a full cutover. But the cost savings and flexibility compound fast. If you're running fewer than 10 simple Zaps, stay on Zapier.