fig4_examples
Two questions from the public test set

The advisor saw only the question. Scores are the share of rubric items satisfied.

“What is Colgate-Palmolive's revenue by geography and product category for each quarter since 2019?”
Advisor's plan
strategydefault
lengthlong (3+ pages)
web useavoid
sourcesFILINGS, …
Watch-outs
  • distinguish metric unavailability from zero
  • enumerate all periods in the requested range
  • derive standalone periods from cumulative filings
  • distinguish explicit from implicit disclosures
What changed

Without advice, the agent filled many quarterly cells with estimates, applying a percentage share to a total. With advice, it laid out the quarterly figures by region and said plainly that Colgate discloses product categories only as a share of sales, not in dollars.

36%
without advice
→
90%
with advice
“What is the summary of ASML's earnings?”
Advisor's plan
strategytables_extraction
lengthmedium (1–3 pages)
web usefallback
sourcesTRANSCRIPTS, FILINGS, NEWS
Watch-outs
  • search across all document types for the topic
  • surface management guidance and forward commentary
  • include geographic and segment breakdowns
  • include capital allocation and investment announcements
What changed

Without advice, the agent centred its summary on the latest quarter and its guidance. With advice, it also built a quarterly grid for the year and added the regional breakdown, including China, and the capital return programme.

47%
without advice
→
100%
with advice

Figure 4: The advisor's plan and its effect on two FrontierFinance queries. Field names are simplified and the tool plan is omitted.