How we work · Founder-led by Shah Durran
We take the recurring manual work off your ecommerce team.
It starts with one question: what does someone still check by hand every day to know the business is running normally? We find that work, scope one focused pilot, and build it — the routine runs on its own, failures and exceptions get surfaced, and the decisions that need a person stay with you.
How an engagement runs
What we build
The work we take off your team.
Examples of the work we take on, not fixed packages. Every engagement is scoped to your tools and your actual bottleneck.
Is the business running normally today?
Daily Operations Monitoring
Scheduled checks across orders, inventory, ad delivery and listings. Missing or stale data and real failures are caught, and a short daily brief tells your team what actually needs attention — not everything, just the exceptions.
Do the numbers actually match?
Reconciliation and Settlement Exceptions
Marketplace settlements, payouts, fees and refunds reconciled on a schedule. What matches is filed; the lines that don't — the exceptions that need a decision — are queued for a person instead of buried in a spreadsheet.
Who handles the routine questions?
Support and Fulfilment Routing
Routine support and fulfilment questions answered or routed from the tools you already use, with anything unusual — refunds, exceptions, edge cases — escalated to your team rather than guessed at.
Ecommerce Operations
Handle the routine. Escalate the exception.
Most operational work is repetition a system can own — order and inventory checks, marketplace reconciliation, the weekly numbers, the first pass on support. We connect the store, the marketplace and your internal tools, let the system handle what's routine, and route only the decisions that need judgment to a person.
Not AI replacing the team. Repetition automated, exceptions surfaced, judgment left to people.
Reference architecture
Shopify + Amazon · illustrativeShopify + Amazon Operations Reference Architecture
A DurranAI-designed workflow showing how an ecommerce team can monitor reporting, inventory, advertising and marketplace exceptions while holding consequential decisions for human approval. Illustrated with synthetic data, not a live client account.
How we work
Decide what to automate. Then build.
We don't automate a process just because we can. First we decide what should run on its own, what stays human, and what should escalate — then we engineer around that.
“What does someone on your team still have to check manually every day just to make sure the business is running normally?”
Whatever the answer is, that's the work we scope first.
- 01
Map the bottleneck
We trace the existing workflow — where people spend time, what data moves through it, and where judgment is actually required.
- 02
Design the system
Triggers, integrations, AI decisions, guardrails, approvals and exception paths, defined before anything is built.
- 03
Build and connect
We implement the workflow, the agents, the application layer and the integrations.
- 04
Test the failure cases
Not just the happy path — what happens when data is missing, an API fails, model confidence is low, or an exception needs a person.
- 05
Ship and iterate
We go live once scope, access and data are ready — typically 1 to 2 weeks after that point — then observe and improve on real usage.
The pilot
One workflow, scoped and bounded.
The first engagement is deliberately small: the one workflow you already know is costing time, built end to end, agreed before anyone starts.
What it includes
- One workflow you already know is costing time, scoped end to end.
- The integrations, checks and exception paths that workflow needs.
- Human-approval and escalation boundaries defined before we build.
- Documentation and a walkthrough with whoever will run it.
What it isn't
- A rebuild of every process at once — one workflow leads.
- Buying tooling you don't need; if an existing tool is enough, we say so.
- Open-ended scope — the pilot is bounded and agreed up front.
How we agree it's done
We agree the acceptance criteria before building: the pilot is done when that workflow runs on its own, surfaces the exceptions we defined, and your team can operate it without us.
When it's live
Live in 1 to 2 weeks after scope, access and data are ready — not before.
What you own
The workflows, prompts and account configuration are yours and leave with you. They run on your infrastructure and your accounts wherever the platform allows it.
Handover and support
Documented and walked through with your team. Beyond the pilot, ongoing monitoring, fixes and iteration are available as a scoped Managed Operations engagement — never a lock-in to start.
Fit
Who this is for, and who it isn't.
Good fit
- Ecommerce brands on Shopify and Amazon.
- Enough order volume that the routine work repeats daily.
- A founder or team still checking and approving things by hand.
- Reporting and reconciliation spread across spreadsheets.
Not a fit
- Pre-launch, or very low volume.
- Looking for a standalone chatbot.
- Wanting AI applied to everything at once, before any diagnosis.
Selected Systems
Built systems, not slide decks.
A build from the operations side: the marketplace back-office system, end to end.
Amazon marketplace reconciliation
Multi-country tax across a large SKU catalog, handled by hand in spreadsheets — slow, and easy to get wrong. We built the system that runs it end to end.
Architecture
Engineering layer
The work leads. The engineering underneath it isn't improvised.
Engineering layer
The work leads. The engineering underneath it isn't improvised.
AI / Intelligence
- AI agents
- LLM integrations
- RAG
- Vision workflows
- Structured outputs
Automation / Orchestration
- n8n
- Make.com
- APIs & webhooks
- Event-driven workflows
Commerce / Integrations
- Shopify
- Amazon SP-API
- Marketplaces
- Messaging
- Reporting & data
Application Layer
- React
- Next.js
- Dashboards
- Internal tools
- Full-stack apps
Engagements
Start with one workflow. Grow into managed operations.
Fixed-price, scoped before anyone starts. Most brands start with a focused pilot; operations can then be run and maintained over time.
Focused Workflow Pilot
From $1,500
Live in 1 to 2 weeks after scope, access and data are ready
One workflow you already know is costing time, scoped, built, and handed back running. The clear first step.
- Reporting flow
- Reconciliation task
- An operational automation
- A focused integration
Managed Operations
Setup + scoped monthly
Operations run and maintained over time: a setup engagement, then a scoped monthly for monitoring, fixes and iteration.
Custom Expansion — Scoped after diagnosis. Larger, multi-system builds, scoped only after a pilot or teardown shows what is worth it. Not an off-the-shelf package.
Questions
Before you book.
- How does an engagement start?
- With a free ecommerce teardown: one call to find the recurring manual work worth automating first. If nothing is worth it, we say so. If it is, we scope a focused pilot before any building starts.
- How much does a project cost?
- A Focused Workflow Pilot starts from $1,500 for one bounded workflow, scoped up front. Ongoing work runs as Managed Operations — a setup engagement plus a scoped monthly. Larger multi-system builds are scoped only after a pilot or teardown shows what is worth it.
- How long until it's live?
- Typically 1 to 2 weeks after scope, access and data are ready — not before. The pilot is bounded, so timing depends on access and clean data more than on build effort.
- What still needs a person?
- The decisions. The system handles the routine, surfaces failures and exceptions, and holds anything consequential for human approval. We define those boundaries before we build, and nothing consequential executes without someone approving it.
- Who actually does the work?
- Shah Durran. Whoever scopes your build is the person who builds it and the person who answers when it escalates — not an account manager and not a roster you're assigned from.
- Do we own what you build?
- Yes. The workflows, prompts and configuration are yours and run on your own accounts and infrastructure wherever the platform allows, so nothing depends on us staying in the room. Handover is documented and walked through with your team.
About the founder
Founder
Shah Durran
Founder, Durran AI
Software engineer, Make.com Certified Specialist and n8n automation specialist, building AI agents, RAG and full-stack systems for ecommerce brands.
“I didn't set out to put AI everywhere. I started by trying to stop doing the same job twice.”
Every useful automation starts with a repetitive process. The goal was never to remove people from the business — it's to remove the repetition, and leave them with the judgment, the exceptions and the relationships.
That's why every system on this page is built the same way: automate the routine, guard the edges, escalate what matters, and keep a person on the decision.
The Teardown
Bring us the part of your operation that eats your team's time.
We'll map the bottleneck, identify what's worth automating, and show you the system we'd build. No obligation.