Deal analysis fast enough to use, clear enough to trust.
Orbe is an AI deal-analysis platform for venture, private equity and public-market investors. I co-designed every flow, from company profiles to an instant LBO model.
Investors judge a deal by stitching together decks, data rooms, news and spreadsheets, often for days, and generic AI doesn’t know their fund.
What I did
Co-owned every flow with one other designer, from spec to handoff: three investment tracks, AI insights, an instant LBO model and the deal flow.
Impact
~90 s
to a first analysis
A first deal analysis in about 90 seconds, with every AI insight linked to its source. 16 design partners on 3 continents; one PE fund became a paying customer in under 30 days.
01
The problem
Private-market investors make high-stakes calls on incomplete, scattered information. For each deal they read pitch decks, dig through data rooms, track news and rebuild financial models by hand. Venture, private equity and public-market teams each do it differently, but every tool on the market treated them the same.
Illustration. Values shown are examples, not product data.
How might we help every type of investor find, understand and model a deal in minutes, without asking them to trust a black box?
02
Who I designed for
Orbe connects three people. Product specs, iterated to v3, defined each hand-off between them.
Fund adminNeeds: set up the fund and control who sees what. Six roles, Admin or Viewer access.
InvestorNeeds: find the right company, trust the analysis, move fast. Invites startups to pitch.
FounderNeeds: share the right data once, securely, without retyping what the investor already entered.
03
Key bets and the options I weighed
Each bet below shows the options on the table and why I chose one.
Bet 1
Serve three kinds of investors without three products
VC, PE and public-market teams start from different questions. How do we serve all three?
A · Three separate productsBest fit per user, but triples design and engineering cost.
B · One generic dashboardCheap, but no one feels it was built for them.
C · One platform, choose a trackChosenTailored views from shared components.
Fig. 1 · Users pick a track once; dashboard, company view and analysis adapt to it.
Bet 2
Help investors decide where to look first
Long lists of companies left screening entirely to the analyst.
A · Sortable table onlyFamiliar, but every judgement stays manual.
B · AI picks in a feedFast, but hides the data investors want to compare.
C · Table plus an ORBE ScoreChosenKeeps the full data and adds a ranked signal.
Fig. 2 · Every row ends in an ORBE Score, turning a list into a shortlist.
Bet 3
Turn hours of LBO modeling into seconds
PE teams rebuild leveraged buyout models in spreadsheets for every deal.
A · Export an Excel templateFamiliar, but still hours of manual work.
B · Static PDF reportFast, but analysts can't test their own assumptions.
C · Instant, editable modelChosenGenerated from company data, with editable inputs.
Fig. 3 · A full buyout model in one view: inputs, returns, sensitivity and projections.
04
The Private Equity flow, end to end
Here is how the bets come together for a PE investor reviewing a private company, from the first dashboard to the files behind the numbers. Step through the shipped screens.
Step 1 of 7Start on the dashboardLatest news, portfolio trends by vertical and every deal in one table, each row ending in an ORBE Score.
Step 2 of 7Pick a trackThe track picker opens over the dashboard, so investors switch context without losing their place.
Step 3 of 7The dashboard re-forms for PESame components, new content: private-equity news and a company table with revenue and ORBE Score.
Step 4 of 7Answer first on the company pageFounding, location, industry and key management up top, then sentiment, deal score and core financials as cards. Regenerate Insights and Create Investment Memo sit right under the summary.
Step 5 of 7Evidence one tab awayMarket size, growth estimate, product, sentiment with recent mentions, and the full income statement, balance sheet and cash flow.
Step 6 of 7Model the buyout instantlyA buyout model generated from the company's data, with editable assumptions, returns, sensitivity and projections.
Step 7 of 7Sources stay with the dealCompany profile, financial model and other documents live in the deal's Files tab, so every number can be traced.
05
Craft: making a dense model readable
The LBO screen holds dozens of numbers. Three details make it scannable.
Fig. 5 · Returns by entry and exit price. The darkest cells meet the fund’s 25% target.
Darker means better, for everyoneAn earlier red-to-green scale failed for color-blind users. One hue in three steps, plus a legend, shows which deals meet the 25% target in any color vision.
Pencils mark inputsOnly assumptions are editable. Everything else is calculated, so nobody breaks the model.
One number highlightedOf every return line, only the final IRR sits in lime.
06
Designing for when the AI isn't sure
Speed only helps if people trust what they see. The shipped screens show the happy path, so I mapped how Orbe should behave when the AI is still working, partly sure, or missing data.
Concept · proposed for the next iteration
1 · WorkingAnalysing 14 documents
✓ Pitch deck read✓ Financial model read○ News and sentiment…
Shows real progress, not a spinner, so 90 seconds feels short.
2 · Partly sureRevenue is growing fast, but costs are rising faster.
Medium confidence3 of 5 sources
See sources →Every claim shows how sure it is and where it came from.
3 · Missing data2023 financials not foundThe model uses estimates for this year until real numbers arrive.
Ask founderUpload file
Gaps become a next step, often a request to the founder's data room.
Illustration: every insight carries its sources and confidence, one click from the exact page or row.
07
Impact
~90sto a first deal analysis
16design partners live
<30days for a PE fund to become a paying customer
3tracks, one design system
"Within 20 minutes of the first demo… this solves the institutional memory gap."
Managing Partner, Deep-Tech VC · quoted on orbe360.ai. Orbe keeps every reviewed deal, so a fund's knowledge stays when people move on.
Mid-market PE fund · USReplaced its manual deal tracking and memo workflow with Orbe, activated 5 licences in 2 weeks and converted to paid within 30 days.Validates Bet 3 and the memo flow
Endowment fund · USJoined as a design partner to shape deal scoring, which now standardizes how its venture team compares opportunities across its full portfolio history.Validates Bet 2, the ORBE Score
Customer results as reported on orbe360.ai. Next, I'd instrument the north-star and guardrail metrics above to measure design impact directly.
08
Reflection
What workedShared components made three tracks feasible for a small team.
What I'd changeDesign the uncertain states from day one, and test the ORBE Score's explainability with investors earlier. A score users can't question is a score they won't trust.
What's nextA mobile companion for partners, and a drill-down showing what moves each score.
