Retail teams are rich in data but poor in throughput.
Category, supply chain, and finance leaders are expected to move faster every quarter. Yet a huge share of their capacity disappears into low-value operational work:
- pulling exports from SAP/ERP
- chasing supplier emails and attachments
- copying values between spreadsheets and portals
- reconciling “almost matching” lists
- following up, escalating, and documenting outcomes
That isn’t strategy. It’s glue work.
And it creates a hidden tax that shows up as lost sales, margin leakage, slow execution, and burned-out teams.
The real bottleneck isn’t insight. It’s unclosed work.
Most retailers don’t have a “data problem.” They have a closure problem.
Work gets stuck between systems, teams, and approvals. Exceptions pile up. Outcomes dissolve into “someone will handle it” instead of “it’s closed, written back, and proven.”
This is the Throughput Problem in retail operations.
The Integration Tax
In most retailers, a single operational decision touches five to ten tools because work is scattered across:
ERP + supplier portals + email + Excel + BI + shared drives + Teams.
Every handoff adds delay. Every context switch adds errors. Every exception creates an open loop.
That is the Integration Tax: the cost of running operations across tools that don’t naturally work together.
It forces people to become the integration layer.
It forces people to become the Human API.
Why most automation stalls after the first 20%
Many retailers tried automation. Most got some wins… and then hit a wall.
Three structural barriers explain why:
- IT dependency
The minute automation needs integrations, access, data cleanup, and long change cycles, business teams stop moving.
- Messy, exception-heavy reality
Supplier formats differ. Data is inconsistent. Portals change layouts. The “happy path” is rare.
- Tools still expect builders
Even “no-code” often means “no-code for engineers.” Category and finance teams shouldn’t need to think like developers to automate work.
So the backlog grows, the easy wins are already taken, and the Human API work stays.
The opportunity: move from assistance to closure
Most AI in enterprise still “helps you work”:
- it drafts
- it summarises
- it suggests next steps
Then a human still has to execute everything across systems.
Retail doesn’t need more suggestions.
Retail needs closed work: executed end-to-end, written back into the right systems, and packaged with proof you can forward.
Our term for today’s reality: the patchwork stack
Retail operations run on a patchwork stack.
Tools were bought at different times, built by different vendors, and connected by… people.
The result isn’t just inefficiency. It’s latency and leakage.
What duvo.ai is building: an AI workforce for real retail work
duvo.ai is an AI agent platform built by retail operators that runs operational workflows end-to-end across the patchwork stack:
- logs into SAP and other ERPs
- works in spreadsheets, shared drives, and email
- handles supplier portals through the browser
- escalates exceptions and requests approvals where needed
- can make outbound calls when the bottleneck is a phone call
- writes outcomes back into systems with full auditability
This is not “automation for automation’s sake.”
Our aim is to be known as:
The system that runs operational work end-to-end with governance, verified write-backs, and forwardable proof.
Governed execution, not black-box AI
Retail teams will not accept “trust me” automation.
So we build control in from day one:
- role-based access and least privilege
- human approvals for high-risk actions
- complete audit trails and job reports
- strict isolation for agent runs in a secure remote browser
Autonomy gives speed.
Governance makes it safe.
You need both.
What changes when you remove the Integration Tax
The value shows up quickly because it returns time and reduces errors immediately.
Examples of what Duvo has already delivered in retail workflows:
- weekly margin and promo reviews standardised, saving ~120 hours per week and giving teams an extra 48 hours before the trading week starts
- multi-market turnover bonus optimisation automated, saving ~250 hours per week and effectively returning 6 FTEs
- key category-management workflows trending toward ~40% effort reduction
That’s not “tasks touched.”
That’s capacity returned, with work closed and written back.
The mission
A mission worth building around has to be measurable.
Here’s ours:
Make repetitive, cross-system operational work in retail self-running - and give teams back at least one day per week, with proof.
Not as a consulting project.
Not as a long integration programme.
As a governed AI workforce that closes cases in your systems, every day.
Martin Pecha
