Product roadmap
Give your storefront
an AI reliability engineer.
We’re building an AI reliability engineer that takes a storefront problem from investigation to resolution—and stays with it to see whether the solution holds.
Shopper behavior, performance and technical evidence. One responsibility: keep the buying experience working.
In the product
Understand the problem.
Follow the result.
Explore shopper journeys, friction, storefront performance and technical failures. Prioritize the issue, give your team the context to act, and track what happens after a change.
Change follow-up is part of Dozenfold today: compare releases, follow issue recurrence and inspect the observed result.
Planned · Deeper commerce understanding
Understand the action.
Not just the click.
Did the coupon apply? Did the cart update? Could the shopper recover and continue? Bring attempts, expected behavior and observed outcomes into the same investigation.
Connect each attempt to the store’s intended behavior and the outcome the shopper actually reached. Distinguish a correct rejection from a technical failure, a successful recovery or an outcome that still needs investigation.
Apply a discount code
- 01
What did the shopper try?
Submit the code, retry, then leave the cart.
- 02
What should have happened?
Check the store’s intended rules. Ask the merchant if eligibility is unclear.
- 03
Which outcome needs attention?
Applied · correctly rejected · technical failure · outcome unknown.
Replay available · Heatmaps planned · Opt-in modules
Session replay and heatmaps.
When you choose them.
Some investigations need to see the page as the shopper did. Error-session replay is available as a separate module that a store turns on deliberately; heatmaps will follow the same way. Neither is part of the default collector.
Inputs stay masked by default, the modules load only where they are enabled, and every replay stays connected to the same issue, journey and revenue evidence already in Dozenfold.
Choose what to capture
- ✓
Core collector · always on
Errors, requests, performance, journeys and funnel evidence. No screens or input values.
- +
Error-session replay · opt-in
Masked page reconstruction of the minute before an error and the rest of that session, linked to the issue.
- +
Heatmaps · opt-in
Click and scroll patterns by page group and device.
Available · MCP & tool access
Your AI tools.
Your storefront context.
Give your coding assistant access to the journeys, performance, issues and change outcomes in Dozenfold. Investigate a real storefront problem with the evidence beside the code.
MCP lets compatible AI tools query a read-only view of one store with a key you create and revoke. Dozenfold’s own agent will use the same commerce context as it works with connected repositories and tools. Connect your AI tools →
“Investigate the cart failures on mobile.”
- ↓
Find the affected flow
Scope the issue, shopper segment and observation window.
- ↔
Read connected evidence
Journeys · requests · performance · source context · releases.
- ↗
Investigate with the code in reach
The assistant receives the evidence and its limits, with references back to Dozenfold.
Planned · AI resolution agent
Give it the problem.
Let it work on the fix.
Dozenfold will decide what to investigate, read relevant code, work out a solution and prepare a tested change. It will act through connected tools within the permissions you give it.
When it needs a business decision or approval, it will ask and resume from the answer. After a change, it will check the result against Dozenfold’s measured comparison—whether the issue stopped recurring and purchases recovered—and keep working when they haven’t.
Stay with the outcome.
Starting with bounded storefront problems and reviewed changes. Broader autonomy grows with demonstrated reliability and the authority you choose to give.
Resolve the failing cart update.
- 01UNDERSTAND THE PROBLEM
A delivery change leaves shoppers stuck.
Inspect affected journeys, failed requests and the release. Establish who is affected and what should happen.
- 02INVESTIGATE
Follow the evidence into the code.
Read the connected repository, trace the delivery handler and test a possible cause against the captured behavior.
- 03WORK ON THE SOLUTION
Prepare the fix and a regression test.
Make the scoped change in a branch. Check the intended cart behavior and explain what the patch changes.
- 04USE THE AGREED PERMISSIONS
Bring the change into your workflow.
Open it for review, ask when a decision is needed, and resume after approval or deployment.
- ↗FOLLOW THROUGH
Check the outcome. Keep working if needed.
Use the change checks and live storefront evidence to assess the result. Continue investigating if the problem persists or the answer is unclear.
Working on your behalf
An agent with a job.
And a clear mandate.
You set its authority.
Choose the stores, code and tools it can access. Start with reviewed changes; production actions stay within the permissions you grant.
It asks when judgment matters.
Store policy, intended behavior and business priorities can require your input. Your answer becomes part of the work.
It stays accountable to the result.
Keep evidence, changes and outcomes connected. Uncertainty stays visible; an inconclusive result is a reason to continue.
Put Dozenfold to work
Start with your storefront.
See the product and the direction we’re building toward.