neo.persona← WORKFLOW LEADERBOARD

PRODUCT TEST / MAKE / 17 NOV 2026

BEST FIRST TRY FOR BEGINNERS

A strong AI start, with connections still to sort out.

Maia built a workflow from our request, and the four-module Canva test completed after a Tavily connection was corrected. The full returned report was not visible in the inspection tool, so we cannot judge every conclusion or citation.

4 / 5
Provisional experience score
A judgment of the observed journey, not a measured product-wide rating.
Best for: A non-programmer or small team that wants to describe a goal and have an assistant assemble a first workflow. Be prepared to connect external services and verify the result.

THE SCORE, EXPLAINED

4/5 reflects the observed ease of beginning and the completed module chain, tempered by external setup and the unverified full output. It is an editorial partial-experience score, not a product-wide reliability rating.

Scale: 1 = severe friction in the observed task; 3 = useful but meaningful setup or verification work; 5 = clear, smooth task experience with a verifiable outcome. We did not measure total completion time or test equal end-to-end scope.

THE JOB

Ask the product's AI to build a workflow that accepts a product name, finds public user reviews, summarizes the main pros and cons and produces a report with source links. We used Canva as the example. The task is a test of Make, not a review of Canva.

WHAT WORKED / WHERE IT WAS HARD

What worked

  • Maia turned a plain-language task into a structured scenario, making it easier to start than assembling every node from scratch.
  • The observed Canva run passed a product name to Tavily, passed search data to a summarizer, and returned a report; the inspection reported four successful modules.
  • In a separate Gmail automation, the user reported that Make understood her request and largely designed the flow for her.

Friction

  • Search needed a separate Tavily account and connection. Previous 401 failures may have been caused by incomplete credential entry by our operator; they are not counted as a Make defect.
  • The assistant inspector displayed only previews of the returned report, even after the output was split. We could not verify all pros, cons and citations.
  • In a separate workflow the user reported that external tool accounts, waiting and a step that stalled without a clear alert made setup less smooth.

THE TEST JOURNEY

  1. Asked Maia to build a product-review search and summary workflow in natural language.
  2. Reviewed its initial proposal and prerequisites. The screenshot shows an earlier Jina search proposal; the final tested search connection used Tavily.
  3. Connected Tavily, corrected an incomplete credential entry, and ran the Canva test. Maia inspection reported five results, HTTP 200 and four completed modules.
  4. Reviewed returned report previews; full output remained unavailable in the assistant inspection view.
Initial Maia plan. This proposal mentions Jina; the run later used Tavily.
Initial Maia plan. This proposal mentions Jina; the run later used Tavily.

THE BIGGEST IMPROVEMENT

Make should carry the user from a generated plan to a visible, verifiable result.

Recommendation: Show which connections are required, make the active run and any stalled step easy to spot, and provide a full output view with source links. Preserve the plain-language workflow builder.

Tradeoff: Setup steps may be necessary for external services. The issue is making that work and the result clear; our credential mistake is not product evidence.

PRICE, CREDITS & TIME

The Make Free plan lists 1,000 credits per month. Before this test the account showed 974/1,000; the final balance was not recorded, so the cost of this task is unknown. Tavily is a separate service. No reliable total task time was recorded.

Make pricing and credits ↗ · Plan information checked 17 November 2026; terms can change.

WHAT THIS TEST DOESN'T PROVE

One controller-led pilot, plus separately attributed real-user feedback on a different task. The full report, citations, repeatability, timing and ongoing cost were not verified. It is not a completed three-persona test.

← BACK TO LEADERBOARDREAD FULL COMPARISON ↗中文版对比 ↗

Evidence-led product testing · This page describes observed actions and limitations. No fictitious persona quotations or unmeasured timings are presented as test results.