共有:

From Research to Deck

The research AI
now builds the deck too.

Research platform AlphaSense has shipped "Work Products," native AI assistants built directly into PowerPoint and Excel. Beyond gathering research, it can now produce sourced pitch decks and financial models without leaving those apps.

AI Navigate Editorial2026.09.226 min read

Research Work Products PowerPoint Excel
01

What Happened

The work "after research"
stays in the same tool

Institutional and corporate research platform AlphaSense announced "Work Products," native AI assistants built into PowerPoint and Excel. From a single prompt, users can generate pitch decks, financial models, memos, and competitive analyses formatted to their firm's branded templates or team standards.

The defining idea is keeping research and deliverable-building inside the same app. Per the July 14, 2026 release, on the Excel side users can pull AlphaSense's financials, filings, and earnings call transcripts directly in, sourcing intact. The design removes the back-and-forth between researching and formatting.

This wasn't a single-feature bolt-on: AlphaSense frames it as part of a broader overhaul that also brings slide-generation workflows into its existing "Generative Search" tool. In effect, AlphaSense is moving beyond its original role of surfacing search results and into assembling the deliverable itself as an extension of that search.

BeforeAfter Work Products
Manually turn research into slidesAuto-generate a draft from one prompt
Fact-check numbers via separate copy-pasteSourced automatically, fully traceable
Apply templates by handKeeps your firm's branding/team format
Research and deck-building are separate stepsBoth happen inside one app
02

What one customer reports

60–70%
Time cut to finish a deck (Blue Deer Capital)
Jul 14 2026
Work Products announcement date

Investment advisory firm Blue Deer Capital says that going from complex research to a board-ready presentation got roughly 60–70% faster after adopting AlphaSense for PowerPoint, according to AlphaSense's own customer profile. That's one firm's account, not a guaranteed average across every customer — worth keeping in mind before extrapolating.


The late nights spent formatting slides
after the research was done —pulled forward and finished early.


03

Who It Affects

Whose overtime this actually cuts

IB / PE analysts

Before a board deck or investment-committee memo is due, you no longer have to run research and formatting in parallel — more time goes into verifying the content itself.

Corporate strategy / IR

Competitive analysis and market summaries can go out fast in your existing internal format, shortening the prep cycle before leadership meetings.

Compliance / audit

Sourcing persists in the output, easing the burden of tracing a number back to its origin — but it adds a new checkpoint: verifying the accuracy of the AI's own summarized wording.

04

Why Now / Risk

The next battleground for "research AI"

Financial and legal research AI has spent the last year or two mostly competing on search and summarization accuracy. AlphaSense pushing into deliverable generation reflects a different insight: research and deck-building had been split across separate tools and workflows, and the seam between them was where the time actually went. A similar move shows up in Kensho's Copilot integration — one more sign that competition in financial AI is shifting from "how fast can you find information" to "how fast can you turn it into a finished deliverable."

Some caution is warranted. The 60–70% figure comes from one customer's account, not a guarantee AlphaSense makes for every user. And shipping an AI-drafted deliverable straight out the door carries a real risk of misreading or miscitation at the summarization step. For high-stakes output — board decks, investment-committee memos especially — a human review pass after generation isn't optional. The efficiency gain and the review overhead it still requires should keep being discussed together, not as a trade one replaces the other. A practical starting point is using it for internal drafts and first passes that already assume human review, with a clear rule against sending generated output straight to external stakeholders unreviewed.