Skip to content

Early access · Interactive preview below

Your AI assistant is only as current as the documents behind it.

Sourcewell scans your source library for outdated policies, duplicate copies and documents that contradict each other. Owners review each finding and approve the fix, before the assistant repeats the wrong answer.

  • A person reviews every flag. Nothing changes on its own.
  • Exact matches and model inferences are always labeled apart.

Findings you can weigh

Exact matches, similarity scores and model suggestions are each labeled for what they are.

Owners and deadlines

Every document has a named owner and a review date, so stale sources surface on schedule.

People decide

Sourcewell proposes. A reviewer approves. It never edits a document or an assistant’s knowledge itself.

A record that holds up

Each decision is logged with the reviewer, the date and the passages they relied on.

The problem

Your assistant can’t tell which policy is the real one.

Support bots, internal copilots and retrieval systems all read from document libraries that people maintain by hand. When those libraries drift, the assistant repeats the drift to customers and employees, confidently.

Retired terms keep answering

A policy changes, the old PDF stays in the library, and retrieval has no way to know which one is real.

Two documents, two answers

A finance policy and a support macro disagree. The assistant quotes whichever ranks higher that day.

Nobody owns the source

Copies multiply across folders with no owner, no review date and no record of what was approved.

Today

Reviews happen by accident.

  • A customer asks for a refund on day 20. The policy says 30 days; the macro says 14. The answer depends on which document the assistant retrieved.
  • Review dates live in a spreadsheet, if they live anywhere. Nobody sees what is overdue.
  • Spot checks cover a few hundred files. Libraries hold thousands, and each rewrite adds more copies.
  • The cost is wrong answers, refunds honored by mistake, escalations, and hours spent working out which document was approved.

With Sourcewell

One queue, one owner per finding.

  • The conflict is flagged with both passages side by side, before a customer hits it.
  • Duplicates and stale files are found by hash, similarity and review date, then routed to an owner.
  • The reviewer picks the current source. A revision task goes to the other document’s owner.
  • The decision is stored with the passages relied on, ready for audits and handovers.

Interactive product preview

Work through a real review, start to finish.

This is a working preview of the Sourcewell workspace, running in your browser on sample data. Open a finding, compare the passages, record a decision and watch the library, tasks and history update.

Fernbrook · Support KnowledgeSample data

Source health

8 open findings · 2 open tasks

Findings by type

Select a bar to open those findings.

open resolved

Patterns in this library

Model-inferred
  • 2 of 3 open potential contradictions are between documents owned by different departments, so no single owner sees both sides.
  • 2 active documents have no owner, so findings on them cannot be routed automatically.
  • M. Okafor owns the most flagged documents (3). Reassigning one or two would balance the review load.

Recent activity

  1. A. Lindqvist: Approved sla-tiers-2026.pdf as current and archived sla-tiers-2023.pdf28 Sep 2026
  2. S. Banerjee: Marked acceptable-use-policy.pdf as reviewed; next review 20 Feb 202722 Sep 2026
  3. M. Okafor: Dismissed possible conflict between billing-faq.docx and invoice-template.docx15 Sep 2026

Fictional company, documents and people. Nothing here is customer data or a measured result, and nothing is saved when you leave the page.

What’s in the first release

Everything you need to keep a source library honest.

Seven capabilities, each one a step in the same review workflow, from the first upload to the signed-off change.

Workflow · Import

Library uploads and scheduled scans

Upload document collections, including scanned PDFs, then re-check them nightly, weekly or on your own schedule.

New and changed files are checked without anyone remembering to ask.

Workflow · Import

Ownership and revision tracking

Every document has a named owner and a full revision trail.

You always know who to ask, and what changed since the last approval.

Workflow · Compare

Duplicate and near-duplicate detection

Identical files are matched by content hash. Passages that say nearly the same thing are scored by embedding similarity.

Copies get retired deliberately, not discovered by a customer.

Workflow · Flag

Potential contradiction flags

A language model suggests passages that may conflict and shows both, side by side, with the exact text highlighted.

Reviewers start from the evidence, not a hunch.

Workflow · Flag

Review deadlines and stale indicators

Set a review cycle per document and see what is overdue before it becomes a wrong answer.

Old guidance is caught on a schedule, not by accident.

Workflow · Review

Assigned review tasks

Route each finding to the right owner with a due date and track it to resolution.

Findings end in action, not in a shared inbox.

Workflow · Approve

Approved-change history

A permanent record of every decision: who approved it, when, and the passages they relied on.

Ready for audits and handovers.

How it works

From a messy library to approved sources, in five steps.

1. Import

Architecture: Amazon S3 · Amazon Cognito

Upload document collections. Every file is stored as a versioned record with an owner and a review date.

  • refunds-policy-v4.pdfFinance Ops · M. OkaforRev 4
  • support-macro-refunds.docxSupport Enablement · J. ReyesRev 2
  • Refunds Policy FINAL (1).pdfNo ownerRev 1

Sample data

How findings are labeled

Certain matches and model judgments are never mixed.

Reviewers should know how much weight a flag deserves. Every finding says how it was found.

Deterministic

Exact duplicates

Identical files are matched by content hash. There is nothing to infer: the bytes match or they don’t.

Method
SHA-256 content hash
Shown as
Match or no match
Similarity score

Near-duplicates

Passages that say almost the same thing are found with embeddings and shown with their score, side by side.

Method
Embedding similarity search
Shown as
Score and aligned passages
Model-inferred

Potential contradictions

A language model suggests passages that may conflict and explains why, citing the exact text. It is a lead, not a verdict.

Method
Grounded model analysis
Shown as
“Potential” with supporting passages

Every flag needs a person to sign off. Sourcewell never changes a document, or an assistant’s knowledge, on its own.

Who it’s for

The people accountable for what the assistant says.

Knowledge managers

The problem
They own the help center and policy library but can’t see which of 4,000 files still match the business.
How they use Sourcewell
One queue of stale, duplicated and conflicting sources, each with an owner and a deadline.
A typical moment
A pricing update ships. They open the stale and conflict findings on pricing documents and send each owner a review task.
The benefit
Fewer outdated answers and a library that stays maintained.

AI engineering teams

The problem
Retrieval quality is limited by the sources. Cleaning them by hand doesn’t scale and nobody records what was approved.
How they use Sourcewell
A reviewed list of which documents are approved to feed the assistant, with duplicates and conflicts surfaced first.
A typical moment
Before re-indexing the assistant, they check that no unresolved contradictions remain in the documents in scope.
The benefit
Cleaner retrieval inputs and a defensible record.

Support operations managers

The problem
Agents and bots both quote macros. When a macro contradicts policy, customers get two answers.
How they use Sourcewell
Side-by-side comparison of macros against policy, with a reviewer decision and a revision task for the macro’s owner.
A typical moment
A 14-day macro is found next to a 30-day policy before an agent, or a bot, quotes it to a customer.
The benefit
Consistent answers and fewer escalations.
Customer industriesSaaSCustomer support outsourcingProfessional services

Technology

How We Use Amazon Web Services

Sourcewell’s job is to turn uploaded documents into findings a reviewer can trust, so its architecture follows that path: files land in Amazon S3, Textract and Lambda prepare them, Bedrock supplies similarity and contradiction analysis, and DynamoDB keeps the findings and approval history that the dashboard reads.

  1. 01

    Data collection

    • Amazon Cognito
    • Amazon API Gateway

    Signed-in users upload document collections through authenticated API calls.

  2. 02

    Secure storage

    • Amazon S3

    Original files and extracted text are kept as versioned objects, one per revision.

  3. 03

    Processing

    • Amazon Textract
    • AWS Lambda
    • Amazon EventBridge

    Scans are read, files are hashed and split into passages, and scheduled scans start on a timer.

  4. 04

    AI intelligence

    • Amazon Bedrock

    Embeddings find near-duplicate passages. A model proposes potential contradictions with cited text.

  5. 05

    Application experience

    • Amazon DynamoDB
    • Amazon CloudWatch

    Findings, owners, tasks and approval history feed the review dashboard. Pipeline health is monitored.

Services and their roles

This describes the architecture Sourcewell is designed around. The interactive preview above runs on sample data in your browser and is not connected to these services.

  • Amazon S3: Provides versioned object storage for uploaded document collections and their extracted text.
  • Amazon Textract: Extracts text and layout from scanned PDFs and images so they can be compared like any other file.
  • AWS Lambda: Runs ingestion, content hashing, passage splitting and the comparison workflow without servers to manage.
  • Amazon Bedrock: Generates embeddings for near-duplicate search and runs grounded analysis that proposes potential contradictions.
  • Amazon DynamoDB: Stores findings, document revisions, owners, review tasks and the approved-change history.
  • Amazon Cognito: Authenticates workspace users and carries their identity into every API request.
  • Amazon API Gateway: Exposes the authenticated API that the review dashboard and uploads call.
  • Amazon EventBridge: Triggers nightly, weekly or custom scan schedules per workspace.
  • Amazon CloudWatch: Collects logs, metrics and alarms for the processing pipeline.

Business progress and early access

Bring one document collection. We’ll help you run the first review.

Sourcewell is at the building stage. We’re talking with teams who maintain a knowledge library behind an AI assistant, and we share what exists today, including what doesn’t.

Company
Privately held · founded 2026
Stage
Building the first release
Product preview
Interactive, on sample data, on this page
Customers
Open to a small number of early-access teams
  • Uploaded collections

    First release
    • Document collections you upload
    • Version comparison and scheduled scans
    • Review queues and assigned tasks
    • Approved-change history
  • Source-system integrations

    After that
    • Connect the systems where documents already live
    • Sync ownership and revisions automatically
  • Publishing to AI systems

    Not available
    • Pushing approved changes to external assistants is not offered yet. Teams export approved changes and apply them themselves.

Request early access

Tell us about your library. We reply from hello@sourcewell.space.

Optional. Tell us about your document library.

We use what you send only to reply and to run the early-access program. See the Privacy Policy.

Pricing

Subscriptions based on workspace and document volume.

Sourcewell is onboarding early-access teams. We confirm plan limits and pricing with each team, so there are no list prices yet.

Starter

Pricing on request

  • 1 workspace
  • Up to 5,000 documents
  • Uploads and weekly scans
  • Duplicate and stale detection
  • Review queue and change history
Request pricing
Most teams

Growth

Pricing on request

  • Up to 5 workspaces
  • Up to 50,000 documents
  • Everything in Starter
  • Daily scans and scanned-document extraction
  • Potential contradiction flags
  • Review deadlines and assigned tasks
Request pricing

Enterprise

Pricing on request

  • Unlimited workspaces
  • Custom document volume
  • Everything in Growth
  • SSO and custom retention
  • Priority access to integrations
Contact sales

Plan scopes show the intended structure and may change during early access.

Company

The Team Behind Sourcewell

Sourcewell is building infrastructure that helps knowledge managers and AI engineering teams keep document libraries clean, reliable and up to date.

  • Samuel Oladipo

    Founder, Sourcewell

    Leads Sourcewell’s product direction, focused on the outdated, duplicated and conflicting source documents that undermine AI assistants.

  • Fidelis Kami

    Co-founder, Sourcewell

    Leads Sourcewell’s technical architecture and AWS design for document processing, embedding, analysis and finding orchestration.

Contact

Questions, pilots, partnerships.

Write to us about early access, pricing, security questions or anything else about Sourcewell.

We use what you send only to reply and to run the early-access program. See the Privacy Policy.