Henry Ingest
We turn scattered and messy data into workable structure.
Your inboxes and drives already hold what you need. They just cannot be queried.
Connect Outlook, Gmail, SharePoint, OneDrive, Google Drive and any other file source. Ingest reads every email and every document, works out what it is, and hands back clean, typed, traceable data.
No migration, no re-filing and no change to how your team works day to day.
Morning,
Please find attached our invoice INV-40881 dated 19 July 2026 for the parts supplied to the Sandton branch last month.
The net amount due is R 48 210.00 excluding VAT, payable within 30 days of statement.
Regards,
Lerato
Extracted
- Supplier
- Delta Parts (Pty) Ltd
- Invoice no
- INV-40881
- Invoice date
- 2026-07-19
- Net amount
- R 48 210.00
- Payment terms
- 30 days from statement
Every value traces back to the words it was read from.
Trust and controls
Reading your mail is a serious permission.
Ingest sits inside your mailboxes and drives. That obliges us to be stricter than a general purpose document tool, not more relaxed.
Scoped, revocable consent
Ingest reads only the mailboxes and folders you nominate, granted by your own administrators and withdrawable at any time.
POPIA aligned processing
Personal information is processed for the purpose you specify, held under a documented retention position and removed on request.
Encrypted, least privilege
Sensitive fields are encrypted at rest, with access restricted to the roles that genuinely need them.
Every read is logged
Access to sensitive fields writes an audit entry, so you can answer who saw what and when without reconstructing it later.
Isolated per tenant
Your data is separated from every other customer at the data layer, not by a filter in the interface.
A person confirms the risky calls
Low confidence extractions never silently become fact. They queue for review with the source alongside the proposed value.
Connect
The mess is already in your mailboxes and drives.
Ingest reads it where it lives. Nothing is migrated, nothing is re-filed, and no team has to change the way it saves or sends anything.
Mailbox
Outlook and Microsoft 365 mailboxes.
Shared, role and individual mailboxes connected through Microsoft Graph. Threads, replies and attachments are read in place, in the order they happened.
- Shared and role mailboxesThe info@, sales@ and accounts@ addresses where unstructured work actually lands.
- Attachments includedEvery attachment unpacked and linked back to the message it arrived on.
- Scoped, revocable consentOnly the mailboxes your administrators nominate, withdrawable at any time.
Personal mailboxes are never connected unless you explicitly nominate them.
- Email and threads
.eml .msg - Documents
.pdf .docx .rtf .txt - Spreadsheets
.xlsx .csv .ods - Scans and photos
.jpg .png .heic .tiff - Voice notes
.mp3 .m4a .ogg - Archives, unpacked
.zip .7z .rar
Email AI
Read the inbox properly, not just faster.
A mailbox is a pile of prose with the facts buried inside it. Ingest reads every thread in Outlook and Gmail, works out what it is about and pulls the specifics out into fields.
- Classification and routingEach message identified against your own categories, then sent to the right team or queue.
- Thread summarisationA forty message back and forth reduced to what was agreed and what is still outstanding.
- Attachment unpackingEvery attachment extracted, classified and linked back to the message it arrived on.
- Entity and reference liftingAccount numbers, order references, dates and amounts pulled out of the prose into typed fields.
- Response draftingA reply drafted from the full thread and the linked record, for a person to check and send.
Document AI
Every file becomes fields you can query.
Whatever lands in SharePoint, OneDrive, Google Drive or an attachment, Ingest identifies it and extracts against a schema built for that type. The output is data, not recognised text.
- Type-aware extractionEach document type has its own schema, so an invoice and a contract are never read the same way.
- Tables as tablesLine items and statement rows come back as rows, ready to sum, not as a paragraph of text.
- Built for bad scansPhone photos, angled scans and low light originals are first class inputs, not edge cases.
- Confidence on every fieldNothing is written as fact without a score, and anything under your threshold is held back.
- Provenance you can proveEach value points back to the exact file and page it was read from.
Sentiment analysis
Tone is data too, and it moves before the numbers do.
Every message carries a signal about how the relationship is going. Ingest scores it, tracks it over time and tells you which accounts are drifting while there is still time to act.
- Per message scoringEach email scored for tone, urgency and escalation risk as it arrives.
- Trend over timeScores roll up per contact, account and team, so a slow slide is visible as a slide.
- Escalation alertsA thread turning sharply negative raises an alert instead of waiting to be noticed.
- Silence countsA conversation that simply stops is flagged, which is often the more expensive signal.
- Reasoned, not just scoredEach score comes with the phrases that drove it, so you can judge it for yourself.
Features
Everything Ingest does, in one list.
Four jobs, in order: reach the data, read it, know how sure you are, and get it into the systems that need it.
Connect
Reach the places your unstructured data already sits.
Outlook and Gmail
Shared, role and individual mailboxes over Microsoft Graph and the Gmail API.
SharePoint and OneDrive
Site libraries, document sets and business drives, watched at folder level.
Google Drive
Shared drives and folders with change-driven pickup.
Other file sources
Object storage, SFTP, network shares, scanner output and direct upload.
Scoped consent
Only what you nominate, granted by your IT team and revocable at any time.
Change feeds
New and updated items arrive as they land rather than on a nightly batch.
Read
Turn each item into fields, not into more text.
Type classification
Every email and file identified against your own taxonomy first.
Field extraction
Typed values with formats normalised across every source.
Table extraction
Line items and statement rows returned as rows, not paragraphs.
OCR and scans
Photographed, angled and low light documents handled as first class inputs.
Attachment handling
Nested attachments unpacked and linked back to the message they arrived on.
Entity resolution
One person or company, even when they appear across many sources.
Thread summaries
A long back and forth reduced to what was agreed and what is outstanding.
Sentiment and tone
Every message scored for tone, urgency and escalation risk.
Duplicate detection
The same document arriving twice is recognised, not double counted.
Trust
Know how sure the system is, and prove where it came from.
Confidence scoring
Every field carries a score, not just a value.
Review queue
Low confidence fields queue for a person with the source alongside.
Configurable thresholds
Set the bar per field and per document type.
Provenance
Every value traces back to the exact email or file and page.
Audit logging
Access to sensitive fields writes an entry you can query later.
Correction feedback
Reviewer decisions improve the classifier over time.
Deliver
Get the structure into the systems that use it.
REST API
Query the structured record directly from your own applications.
Signed webhooks
Push events downstream the moment an item is settled.
System write-back
Results land in your CRM, ERP or line of business system.
Scheduled export
Bulk delivery for reporting and warehouse loads.
Routing rules
Send each type to the right team, queue or system automatically.
Next actions
What is missing, who owns it and what has gone quiet.
How it works
Scattered in. Workable structure out.
Five stages run on every email and every file Ingest receives. Select one to see inside it.
Connect the mailboxes and drives you already use.
Outlook, Gmail, SharePoint, OneDrive, Google Drive and other file sources are connected once, with scoped consent your IT team grants and can withdraw. Nothing moves, nothing is re-filed, and nobody changes how they work.
Scoped access
Only the mailboxes and folders you nominate, never a blanket tenant grant.
Change driven
New and updated items are picked up as they land, not on a nightly sweep.
No forwarding
Nobody has to send anything to a special address for it to be captured.
Backfill optional
Start from today, or reach back through history when you want the archive too.
The output
This is the structure you get back.
One record per entity, assembled from every source that mentioned it, with each field traceable to the email or file it came from.
Delta Parts (Pty) Ltd
Account 40881, owned by FinanceReady to actionStructured fields
Extracted, scored and normalised
- Entity
- Delta Parts (Pty) Ltdjoined from 3 mailboxes and 2 drives
- Registration
- 2018/114872/07from a contract in SharePoint
- Open balance
- R 128 940.00summed from 4 extracted statements
- Latest invoice
- INV-40881, R 55 441.50from a Gmail attachment
- Payment terms
- 30 days from statementfrom the signed supplier agreement
- Tone trend
- Declining, 4 weeksfrom 26 scored messages
Source items
Classified and linked automatically
- Supplier agreementPDF, SharePoint / ContractsVerified
- Statement, JulyPDF, Outlook attachmentVerified
- Invoice INV-40881PDF, Gmail attachmentVerified
- Delivery note, photographedJPG, OneDrive, low lightNeeds review
- Account query threadEmail, 26 messagesVerified
- Proof of paymentexpected, not yet receivedOutstanding
Next actions
Raised from what is missing
Get started
See it on your own mess.
Tell us where your unstructured data actually sits today and we will walk you through Ingest against a real, anonymised example from your world rather than a generic demo.