Finance and shared services

Start with one process. Prove it under audit.

We redesign the close, payables and shared-services work around AI agents, one process at a time, with the controls and evidence your controller and auditors will test.

Each engagement covers one entity, one country or one team.

You see agents working on your own data, with every check logged, before you decide what to scale.

Who it's for

For the people who run the close, payables and the service center.

Record-to-report, procure-to-pay and order-to-cash, in-house or in a shared-services center.

In VirtuBrain

15

Finance processes, each with notes for 6 industries and 3 countries, typical systems and KPIs.

Explore finance

Every check logged.

Each agent action leaves evidence a reviewer can follow.

Agentic Close Assurance

100 days

One entity, one close, from zone map to handover.

See the offer

Accountability

Sign-off stays with a named person.

Agents prepare and execute. The controller still signs the close.

Calibrated

For the Philippines, the United States and Singapore.

Every process carries a note for each country, on filing deadlines and control rules, from BIR VAT returns to SOX 404 and SFRS(I).

Offers

Four ways to start.

Each one is scoped in writing before work starts and ends with work your team keeps. We suggest starting with the close.

Featured100 daysOne entity

Agentic Close Assurance

We map your close into zones, put agents on the reconciliations and matching that can be checked against records, and build the evidence trail your controller and auditors will test.

You get

  1. A zone map of your close
  2. Agent-run reconciliations with logged checks
  3. A control matrix ready for audit testing
  4. A runbook and handover to your team

Measured against KPIs you agree at the start, like the ones VirtuBrain lists for the close.

  • Days to close
  • Journal entries matched without a touch
  • Late adjusting entries after close

This is not an audit or assurance engagement. Your auditors test the controls; we build them and the evidence.

  • Pilot

    Agentic Supplier Invoice and Payment

    Invoice intake to payment proposal for one country or business unit. Agents capture, code and match. Exceptions go to analysts with the reason attached.

    You get

    • A country exception inventory
    • Matching and tolerance rules
    • An exception queue design
    • Before-and-after measures from your own data
  • Framework

    Agent Governance Design

    For teams with agents or copilots already in production. Autonomy is set by how verifiable the work is, and each agent action leaves defined evidence.

    You get

    • Autonomy tiers by process step
    • An evidence and logging standard
    • Roles for review and override
    • A pack internal audit can work from
  • Training

    Team Enablement

    Practical AI fluency for finance, HR and shared-services teams. People practice on their own work, in tiers, with coaching on what they actually write.

    You get

    • A tiered curriculum by role
    • Hands-on practice with feedback
    • A leader briefing on agentic operating models

Approach

How an engagement runs.

Four moves, in order. The same sequence works for a global close or a single payables team.

  1. 01

    Calibrate

    We lay your operating model against the VirtuBrain taxonomy for your industry and country, then agree the scope, down to the entity, the process and the steps in play.

    OutputA scoped process map, step by step

  2. 02

    Diagnose

    We walk the process with the people who run it and find the steps where work stalls. Every finding is tied to its step, with evidence from your own data.

    OutputFindings pinned to numbered steps

  3. 03

    Redesign

    For each step we decide who runs it, whether an agent alone, an agent that drafts for approval, or people. Then we design the controls and the evidence each step leaves.

    OutputA zone map, controls and an evidence standard

  4. 04

    Run

    We stand up the new model on the agreed slice and measure it against the KPIs set at the start. You decide what to scale from the results.

    OutputA working model and before-and-after measures

The three zones

Sort each step by how its output can be checked.

Autonomy follows verifiability. Where the output can be checked against a record, an agent runs it and logs the check. Where it can't, a person stays in the loop.

X-ray every step of the close

Auto

Output can be checked against a record. An agent runs it and the check is logged.

  • Bank reconciliationCheck: the statement ties to the ledger, item by item
  • Intercompany matchingCheck: both sides agree to the cent
  • Recurring journalsCheck: the entry matches the approved schedule

Assist

Output needs judgment to confirm. An agent drafts and a person approves.

  • AccrualsThe agent drafts from open POs, an accountant reviews
  • Flux analysisThe agent explains variances, the controller confirms
  • Balance sheet supportThe agent assembles, a reviewer signs

Human

Accountability can't be delegated. People decide and agents prepare.

  • Judgmental provisionsPolicy and judgment, documented by a person
  • Close sign-offThe controller owns the numbers
  • Auditor conversationsRelationships and accountability

Auto

Output can be checked against a record. An agent runs it and the check is logged.

  • Invoice capture and codingCheck: fields match the source document
  • Three-way matchCheck: PO, receipt and invoice agree within tolerance
  • Duplicate detectionCheck: no prior invoice with the same supplier, number and amount

Assist

Output needs judgment to confirm. An agent drafts and a person approves.

  • Exception resolutionThe agent proposes the fix, an analyst approves
  • Supplier queriesThe agent drafts from ledger status, an analyst sends
  • Tax and country rulesThe agent flags the rule, a specialist confirms

Human

Accountability can't be delegated. People decide and agents prepare.

  • New supplier approvalDue diligence and judgment
  • Payment releaseAn authorized approver, with duties kept separate
  • Dispute settlementA commercial decision

An example zone map. Yours comes out of the diagnostic.

Already mapped

Finance, already broken into 15 processes.

Calibration starts from a map that already exists. Open any finance process to see its industry and country notes, typical systems, KPIs and its breakdown into activities, tasks and steps.

Finance and accounting

15processes in 5 groups
90industry notes, 6 for each process
45country notes, for the Philippines, the United States and Singapore

In the close, one task (8.1.1.2.2, review, approve and post journal entries) carries curated step detail. Every other finance task follows a standard five-step pattern: confirm the inputs, do the task, check the output, decide whether it is complete, then log and hand off. An engagement replaces that pattern with your own steps.

Open finance in the app

See 33 illustrative failure points in financeIllustrativeModeled on process-mining conventions. In an engagement the library is filled with your own evidence.

Principles

What we hold ourselves to.

Three of the six rules that shape every engagement, from the first call to the handover.

  • Automate what can be checked.

    An agent runs a step on its own only when its output can be checked against a record.

  • People stay accountable.

    Agents prepare and execute. Sign-off stays with a named person.

  • No promised percentages.

    We don't quote improvement figures until we have measured your process with your data.

All six principles

FAQ

Questions finance leaders ask first.

Something else on your mind? Ask us directly

With one process in one bounded slice, such as one entity, one country or one team. We usually suggest the close or supplier invoices, because both have many steps whose output can be checked against a record.

By how its work can be checked. If a step's output can be compared with a record, such as a statement, a purchase order or an approved schedule, an agent can run it and log the check. If confirming it needs judgment, the agent drafts and a person approves. If accountability can't be delegated, people decide.

Each agent action leaves defined evidence, recording what it checked, against which record, and the result. The control matrix maps those checks to your controls, so internal and external audit can test them the way they test any other control.

No. We don't quote improvement figures until we have measured your process with your data. The before-and-after measures you get come from your own systems, against KPIs we agree at the start.

We agree data access, where agents run and how evidence is stored in the written scope, before anyone touches your systems.

We quote in writing after the first call, against the scope we agree. No quote depends on promised savings.

The work is designed around the systems you already run. VirtuBrain lists the typical systems for each process. For the close that includes BlackLine, SAP S/4HANA, Oracle Fusion Cloud ERP and Workiva.

The Failure-point library is illustrative. Its failure points were modeled on process-mining conventions against the real taxonomy, so they show the kind of failure to look for and the step it sits on. In an engagement, the diagnostic replaces them with findings from your own evidence.

Verantis is founded and led by Ver Villarico, a Consulting Director at Accenture with a background in record-to-report, procure-to-pay, order-to-cash and shared-services delivery. We're based in Manila and work with clients in the Philippines, the United States and Singapore. More about the practice

Start with your close.

Tell us about the entity and the close you'd like to prove first. We'll reply by email to set up an intro call.