Find trusted answers in your company documents.

Ask a question in plain language and get an answer from SharePoint, SAP, Google Drive and your file shares, with the exact page or cell behind every figure.

See how it works

Self-hosted · Runs on Postgres · Backed by Y Combinator

EigenSearch
How did Retail EBITDA land against budget in June, and why?

4 sources found· 1 locked

  • 1Mgmt report
  • 2Budget 2026
  • 3Retail actuals
  • 4Challenge notes

Retail EBITDA closed June at €12.6m1, €0.8m below the €13.4m budget2. Most of the gap is pharmacy retail, where gross margin fell to 27.9% against 29.6% planned3. Management attributes the drop to promotional pricing on over-the-counter lines4.

Retail EBITDA, June€m
Budget13.42
Actual12.61

−0.8 against budget

Checked against Maria's access. One match she can't open was left out.

Connects to systems your company already uses

See the Knowledge API
  • SharePoint
  • OneDrive
  • Microsoft Teams
  • Outlook
  • Excel
  • Google Drive
  • Gmail
  • Google Sheets
  • Box
  • Dropbox
  • Confluence
  • Notion
  • Slack
  • DocuSign
  • OpenText
  • File shares
  • SAP
  • Oracle
  • Dynamics 365
  • Workday
  • Salesforce
  • ServiceNow
  • Jedox
  • Power BI
  • Tableau
  • Snowflake
  • Databricks
  • Jira
  • Zendesk
  • HubSpot
  • Amazon S3
  • Azure
  • PostgreSQL
  • GitHub
  • Knowledge API

How EigenSearch builds the answers from your sources.

  1. 01 Connect

    Read-only connectors to the systems you already run. Nothing is moved or changed.

  2. 02 Index once

    Every document is read once, split into passages and tagged with entity, period and access group.

  3. 03 Search

    Each question is matched on exact terms and codes, and on meaning. The best passages win, not whole files.

  4. 04 Check access

    Passages the asker may not open are dropped before the AI sees them, titles included.

  5. 05 Answer, cited

    The AI writes only from what's left. Every figure is cited, and gaps are called out.

AI chatbots read what you upload.
EigenSearch searches everything.

JSJulia Stein, CFORetail EBITDA, JuneMon 09:10

How did Retail EBITDA land against budget in June, and why?

Reply
Everyday AI Chatbot

3 files

in this chat

21,315

documents it can't see

MMR_June_draft.pdfBudget_v6.xlsxSales_June.csv

How did Retail EBITDA land against budget in June, and why?

Based on the files you shared, Retail EBITDA came in at €12.9m, about €0.3m under budget. Most of the gap is probably pharmacy.

Ask anything

JSJulia Stein, CFORetail EBITDA, JuneMon 09:10

How did Retail EBITDA land against budget in June, and why?

Reply
RE: Retail EBITDA, JuneTo: Julia SteinSent Mon 09:20

Retail EBITDA came in at €12.9m final report says €12.6m, about €0.3m under budget old budget, v6. Most of the gap is probably pharmacy no source.

Sent with errors, and no source anyone can check

Time to answer

10 min

Cost to answer

€13

Your documents never leave your network.

  • No training on your data

    Your documents and questions are never used to train any model, ours or anyone else's.

  • Runs on your infrastructure

    Documents, index and answers stay on your own servers or in your cloud account.

  • Full audit logs

    Review every question, every answer and every source opened.

Answers follow your existing permissions.

People sign in with their existing account. Their groups decide which documents appear in search and in answers, so access follows your directory.

Works with your identity provider

  • Microsoft Entra ID
  • Google Workspace
  • Okta

and any OIDC or SAML provider

Access example

ANAnna NovákGroups: Finance · All staffEntra ID
  • Management report, JuneFinance
  • Salary review 2026 (locked)HR only
  • Travel and expense policyAll staff
  • Project Falcon, deal memo (locked)Deal team only

2 of 4 documents visible to Anna

Locked documents never appear in search or answers, not even by title.

Control your rollout from one console.

Manage sources, users, sign-on, AI settings and usage in one place. See which systems are connected, when they last synced, and who can search what.

EigenSearch Admin

Sources

4 connected
Add
  • SharePointSynced 4 min ago12,480 documents · read-only
  • SAPSynced 11 min ago3,126 documents · read-only
  • JedoxSyncing1,904 documents · read-only
  • File sharesSynced 1 h ago3,808 documents · read-only
21,318
documents indexed
4
sources
06:00
last full sync

Users & groups

Synced from Entra ID
  • Finance48 membersSharePoint · Jedox · File shares
  • Controlling22 membersSharePoint · SAP · Jedox
  • HR9 membersSharePoint, HR site only
  • Board7 membersAll sources

Groups and members come from your directory. Change them there, and search access follows.

Sign-on

Microsoft Entra IDSAML 2.0 · SCIM provisioningConnected
  • Require sign-on for every user
  • Provision users and groups with SCIM
  • Remove access when someone leaves the directoryApplied at the next sync
  • Allow guest accounts

AI settings

  • Language modelPick the model and hosting that fit your data rulesIn your cloud account
  • EmbeddingsSelf-hosted
  • Cite the page or cell behind every figure
  • When the documents don't cover a questionSay it's not covered
  • Use questions to train models

Usage

Last 30 days
1,284
questions
212
people asking
37
not covered
Questions per day43 a day on average
30 days agoToday

Fits in between the systems you already run.

EigenSearch sits inside your network, between the systems that hold your documents and the people who need answers. Nothing is moved, and your sign-on decides who sees what.

Your cloud or servers

Your systems

  • SharePoint
  • SAP
  • Jedox
  • File shares

Read only. Nothing is moved or changed.

EigenSearchSearch and cited answersAdded

Your people

  • ANMKTBFinance
  • LPRHControlling
  • JSLeadership

Signed in with your identity provider.

  • Private by default

    EigenSearch runs inside your network, so sensitive documents stay within your environment.

  • Your cloud or your servers

    Deploy in your own cloud account or in your data centre, with the same security and access model.

  • Efficient at scale

    Large document libraries are searched without heavy extra storage or memory, so it stays affordable as you grow.

See every figure traced back to its source.

Book a demo to watch a finance question go from search to a cited answer, along with the permissions and deployment your IT team will ask about.

About EigenPal

Part of the EigenPal platform · Backed by Y Combinator