Agents
Exclude wasted search terms, deliver the Monday report, update the logistics sheet, send segments to the ad platforms. Describe the workflow in plain language — an agent runs it on schedule against live data and executes with your approval.
252 live agents across 51 commerce brands
Review search terms weekly. Draft negative keywords for terms that spend without converting, and send them for approval.
Search-term waste negator
Weekly · Mondays
Executive KPI brief
Daily · 07:30
Warehouse order forecast
Weekly · Mondays
Margin feed segments
Daily · 05:00
Trading report
#trading · Slack
KPI brief
Leadership · Email
Order forecast
Logistics sheet
12 negative keywords
Google Ads
Trusted by commercial teams at leading brands
Sound familiar?
Most commercial routines are the same steps on the same schedule — but they run on someone's morning, a dashboard nobody opened, or an AI that can't see your systems.
Build the Monday report
07:00 – 10:30
Repeats weekly · no end date
3.5 hours of someone’s week, every week
Refund rate spiked in Outerwear
Someone finally opened the report
Watch our inventory and alert me when a SKU drops below 4 weeks of cover
I don’t have access to your inventory or order history, but margin was likely around 24% — you should be fine!
✕ It can’t see your systems
✕ It can’t watch anything, or act on anything
How it works in Dema
An agent is a described job with a schedule, a data foundation, and permissions — not a flowchart you maintain. Every example below runs that loop, and links to its full use case.
Profit Pilot
Runs Fridays 16:00
Prepare next week's budget moves where profit is incremental, and send them for approval.
Reasoning
“Reported ROAS says scale everything — but brand search is soaking up credit. Check incrementality before moving budget.”
Queried modeled data
Spend, contribution margin, and returns by channel and market · last 28 days
Asked the MMM
Diminishing-returns curves per channel · profit-optimal split for next week
Used skill: budget-move-guardrails
Max ±20% per move, never against low stock cover
Modeled next week's allocation
3 moves · +€19.5k projected weekly profit
3 budget moves prepared
Awaiting approvalMeta · Google Ads
Plan posted to #growth with the numbers behind each move
From months to one command
“What took two people three to four months now runs in a single command. We built 20 workflows in three weeks without writing a line of code.”
Why it works
An agent is only useful if the numbers are right and the workflow survives your organisation — new joiners, changed owners, and all.
#tradingDema APP
Weekly trading report — DE margin −2.1pts, the three largest movements explained below.
Share it, set who can edit, and have it delivered into Slack on a schedule. When that person leaves, the workflow stays.
budget-move-guardrails
Runs identically for all 8 people
CM3 = Revenue − COGS − fulfilment − marketing
Defined by your team — not inferred
Orders, inventory, returns, marketing and omnichannel data connect once and stay maintained. Nobody on your team owns the pipeline.
And dozens more via APIs and direct integrations.
View all integrationsWhat you can set up
Every card is a real setup from the prompt library. Open the use case behind it to see the full run — what it checks, what comes back, and where it lands.
One foundation, three ways to work
Chat, Agents, and Apps all run on the same modeled commercial data — the answer, the workflow, and the application never disagree.
Which products drove margin last week?
Answer with 3 charts
12s
Ask why margin moved, which products slowed down, or where paid growth stopped being profitable — and get the number that is actually right.
Explore Chat
Sell-through planner
Team app
Reorder Olive Overshirt
240 units
Markdown Rust Tee
−15%
Publishes to the storefront
PublishPlanners, dashboards, review queues — describe the workflow and get a working application on live data, shared with the team.
Explore Apps
Enterprise-ready
Where your data is processed, how it is protected, and what an agent is allowed to change are controls you set — not defaults you inherit.
Choose where your data is processed. Model inference and data pipelines run on European or US infrastructure with full residency control.
Your data is never used to train models. Processing stays in your chosen region and runs in isolated, encrypted sessions.
New reasoning models roll out as they ship, so your agents keep improving automatically without migration work.
Each agent has its own scope and integrations. Actions can require a named approval before anything changes.
Questions
They can remain read-only, draft actions, or execute through connected systems. You decide which tools each agent can use, which actions need approval, and who can approve them.
Recurring commercial work with a clear trigger, decision, and destination: weekly reports, keyword exclusions, budget guardrails, feed segments, sheet syncs, anomaly response, inventory monitoring, CRM activation, and more.
A schedule (every Monday at 07:00, daily at 06:00) or a condition in the data (refund rate breaks its range, a product drops below three weeks of cover). Both run against live modeled data.
Reports and alerts deliver to Slack, email, and Google Sheets. Executed changes go to connected systems — Meta, Google Ads, product feeds, the ERP, and other integrations — always within the permissions you set.
The agent flags it instead of quietly publishing. Missing sources, broken definitions, and failed runs surface to a named owner — the opposite of a zap that fails silently.
Teams connect their core commerce systems first, then begin with one recurring workflow — most start with a weekly report or a spend guardrail and expand from there.
Bring one commercial job
