Enterprise data intake

One controlled intake foundation across email, shared folders, uploads, SFTP and system exports

• Domain: Finance & Accounting
• Moment: Before the work starts
• Customer: An enterprise services company running a major ERP

Bring every intake channel into one controlled process.

The situation

An enterprise services company runs a major ERP as its system of record. The ERP is not the problem. It records correctly what it is given.

The problem sits upstream. Documents reach it through email, shared folders, user uploads, SFTP from counterparties and exports from other systems. Each grew separately, around a team and a deadline, with its own folder convention, its own spreadsheet, its own rules and its own person who knows what to do when a file looks unusual.

Five channels, one ERP, and five definitions of what a valid document looks like.

What breaks today

The channel decides the process. The same counterparty document is handled one way when it is emailed and another when it lands on SFTP. Two paths, two rule sets, two ways to drift apart, for no reason beyond history.

There is no single arrival record. Nobody can answer, at any moment, what arrived today, what is waiting and what was rejected. Each channel answers only for itself, and only while the person who runs it is at their desk.

Validation happens last. Errors surface when the ERP rejects the load, or when a reviewer notices, which is after the keying and often after the period. The cheapest moment to catch a bad field is when it arrives, and nothing checks then.

Reference data is applied from memory. Counterparty identity, entity, cost centre, currency, tax treatment, account mapping. Each channel owner keeps a working list. The lists were identical once.

Exceptions have no home. A file that cannot be processed stays in a mailbox or a folder. It has no queue, no owner and no age, so it is rediscovered when the counterparty chases.

Every new source is a project. A new counterparty, format, or export creates another process for one team. The work cannot reuse a shared intake foundation.

What Manuel does

One intake layer across every channel. Email, shared folders, user upload, SFTP, and system exports enter the same path. Mailbox access uses delegated permissions for named users. A runtime agent can connect to sources behind the firewall.

Registration before reading. Every arrival is recorded with source, channel, sender, timestamp and file fingerprint. A document emailed and then uploaded is caught here, because the fingerprint does not care how the file travelled.

Classification and extraction. One file may hold several business records, and one record may span many pages. Manuel assembles each record and defines its fields once across all channels. A new counterparty layout is added to the existing process.

Enrichment against your own master data. Counterparty resolved on the registered identifier first with name as a fallback, then entity, cost centre, currency, tax treatment and account mapping from the ERP's own records. An unresolved item names its candidates.

Validation before load. Manuel checks required fields, formats, cross-field arithmetic, master data references, and duplicates across the population. A failed document is held with the reason.

One exception queue. Every held item carries a named reason, an owner, a severity and an age.

ERP-ready output with a full history. Every material action retains the actor, timestamp, reason, states before and after, and linked records. A posted row can be traced back through the business record, document, and channel.

What the customer gets

One answer to what arrived, what is waiting and what was rejected, across all five channels, at any hour.

A new source is added through configuration. Later business records reuse the intake, identity matching, validation, and exception handling built for the first.

We have no published throughput result for this deployment. What we baseline in discovery: documents a month by channel, distinct formats per channel, minutes per document, the share reaching the ERP without a manual touch, and days from arrival to posting.

What stays with your existing systems

The ERP stays the system of record with its current posting process. Master data stays where it is maintained, and Manuel reads it when needed.

The mailboxes, folders and SFTP endpoints stay where your counterparties already send. Nobody is asked to change how they submit.

The first deployment produces a validated file the ERP imports through its own path. Write-back is closed afterwards, once permissions, maker-checker and cutover are agreed.

What a first scope looks like

Start with one business record across two channels and a population you have already processed. The read-only test returns a file.

Later channels reuse the agreed record structure, owners, approvals, and evidence rules. Each addition is tested on your own documents before production.

Related: AP invoice control / Freight document intake / Management reporting certification