Company profile · 4 min read
Melder: AI analysis inside Excel
Melder adds AI formulas, document extraction and workbook assistance to Excel for finance, operations and research teams.
Cofounder of Lazyweb (a16z Speedrun). Previously grew subscriptions and built native ads at Duolingo. Writes First1000 (90,000+ subscribers).
Published · Updated
What it does
Melder puts AI document handling and spreadsheet work inside Microsoft Excel. Users can upload a PDF into a cell, call functions such as =GEN() and =EXTRACT(), ask a workbook chat assistant for help, and build charts or cleaned tables without moving the job into a separate AI app. The YC profile names diligence, contract review, competitor research and transcript synthesis as intended business workflows.
That makes the best initial buyer a finance, operations or research team already committed to Excel and allergic to another workflow migration. A team can keep its model, formulas and review habits while adding extraction and classification around the edges. The company’s own launch description emphasizes one-click installation and deployment through Office 365, but the important question is whether the result remains auditable to the person who owns the spreadsheet.
Why I’d look closer
Melder’s wedge is not “AI for spreadsheets” in the abstract. It is the combination of structured spreadsheet logic with unstructured source material: an invoice, contract, report or website can become an input to the existing workbook. The homepage shows separate workflows for financial analysis, sales dashboards, accounting summaries, statistical analysis, consolidation and document review. That breadth is useful for discovery, but I would start with one repeatable task—such as turning a data room into a diligence checklist—rather than hand it an entire finance function.
The pricing page lists Pro at $50/month with a seven-day free trial, Ultra at $200/month, and Team as contact-led. Pro and Ultra differ mainly in operation, formula-credit, file-upload and web-scraping limits; Ultra advertises unlimited formula credits and file uploads. The trial requires no credit card. That is enough to run a contained evaluation before negotiating team terms.
Security claims are clearly vendor claims, not a certification. Melder’s pricing FAQ says spreadsheet data stays in the user’s workbook, that communications use TLS, and that content is not sold, shared or used to train external models. It also says the product follows security practices including permission scopes, code reviews and regular penetration testing. I would verify those boundaries against the exact Excel permissions and the behavior of web-scraping or document operations before putting confidential deal data through it.
The YC founder biographies describe Ashwin Dharne as former Google security and cloud conversational-AI staff, and Philip Brewer as a former Bain consultant building “Excel for unstructured data.” That pairing fits the product’s technical-and-workflow angle.
What I’d ask
What leaves the workbook during a formula call, how are citations or source cells preserved, and what happens when a page or PDF cannot be parsed? I’d test permission scopes, model-change behavior, formula reproducibility and failure handling on a redacted contract set. Team buyers also need to know whether shared workbooks, billing and admin controls match their existing Microsoft governance.
My editorial take
Melder is a credible shortlist for Excel-heavy teams that want AI assistance without abandoning the spreadsheet as the system of record. Its success depends less on flashy generated formulas than on traceability when a reviewer asks, “Where did this cell come from?”
Quick facts
| Field | Sourced detail |
|---|---|
| Buyer fit | Finance, operations and research teams working in Excel |
| Public pricing | Pro $50/month; Ultra $200/month; Team contact-led |
| Security posture | Vendor states workbook-local data handling and TLS; no independent audit claim used |
| Founder context | Former Google security/conversational AI and Bain backgrounds |
Sources checked
Checked 2026-09-19.
| Source | Used for |
|---|---|
| YC company profile | Product workflows, founders and launch description |
| Melder homepage | Current Excel-agent and analysis surface |
| Melder pricing and FAQ | Plans, trial terms and vendor-stated security/data handling |
Cohort context
Melder is listed in Fall 2024. In our 2026-09-18 directory snapshot, 57 of 94 listed companies in that cohort have YC’s primary industry label B2B (60.6%). This is a current-directory comparison, not an original intake count or a performance ranking. Nine-cohort dataset.
Public website snapshot
Observed 2026-09-19T16:14:40.867Z in raw homepage HTML. This records visible metadata and advertised links, not agent execution or product quality.
| Signal | Homepage observation |
|---|---|
| Product description metadata | Observed |
| Canonical link | Observed |
| H1 or H2 heading | Observed |
| Typed structured data | Not observed in this response |
| Docs/developer link | Not observed in this response |
| Pricing link | Observed |
| llms.txt link | Not observed in this response |
| Markdown alternate | Not observed in this response |
Public observations · Collection method. Missing links here do not establish that a capability or file is absent elsewhere.
About the author
Cofounder of Lazyweb (a16z Speedrun). Previously grew subscriptions and built native ads at Duolingo. Writes First1000 (90,000+ subscribers).
I cofound Lazyweb and publish Mudpie. This is an owner-written publication, not an independent testing organization. Research notes distinguish observations, sourced reporting and editorial judgment.
First1000 ↗ · X ↗