How Tatti Lashes Built an AI-Powered Planning System in Three Weeks
Tatti Lashes is a UK-based beauty brand specialising in premium false eyelashes, cluster lashes, and lash accessories. Known for their viral Invisi-lash kits and a loyal following among makeup artists and beauty enthusiasts, the brand sells direct-to-consumer via Shopify and through TikTok Shop, serving customers across the UK, Ireland, and internationally. As the brand prepared for its next financial year, the team faced a growing operational challenge: turning months of sales history, product-level data, and marketing performance into actionable procurement and budget plans without drowning in spreadsheets. That's when they turned to Dema's commerce AI agent and began building custom skills to automate their most complex planning workflows.
The Challenge
Tatti Lashes' team was planning across multiple dimensions simultaneously: revenue forecasting, procurement, marketing budgets, and channel reporting. Each required pulling data from Shopify, ad platforms, and finance files, then layering assumptions, scenarios, and strategic overlays on top. Three pain points stood out:
What's on your mind today?
Procurement planning complexity
With hundreds of SKUs across multiple sales channels, building a bottom-up procurement forecast meant working at the product × channel level across 12 months, factoring in seasonality, declining categories, new product launches, and brand collaborations.
Scenario modelling
The business needed to model multiple futures, from baseline projections to CRM-driven uplifts, new product launches with cannibalization estimates, and celebrity collaborations, each producing finance-ready output in a consistent format.
Recurring reporting
Weekly paid social reports, market-level performance analysis, TikTok Shop analytics, and monthly business updates each required a specific methodology, data sources, and output format, repeated on a weekly or monthly cadence.
The Solution
Over the course of three weeks, the Tatti Lashes team built 20 custom skills for their Dema commerce AI agent, creating what is effectively an AI-powered planning and reporting operating system for the brand. Critically, the team didn't write any code or prompts themselves. They simply talked to the agent, describing the problem they wanted to solve, the data they needed, and the output they expected. The agent then wrote the skills and Python code required to execute the workflow, and the team could iterate on the results through conversation until the output matched their requirements. Once satisfied, they saved the workflow as a reusable skill, turning a one-off conversation into a repeatable, deterministic process that anyone on the team can trigger with a single command and get the same structured output every time.
Skills
Skills guide the agent for specific types of analysis. Triggered automatically or manually with /[skill name].
Weekly Mix Diagnostic
Analyze channel spend vs CM3 across all markets with YoY comparison
Added by you · 2d ago
Margin Alert
Flag campaigns where ROAS dropped below threshold in the last 7 days
Added by you · 5d ago
Stockout Detector
Cross-reference inventory forecasts with active ad spend and suggest exclusions
Added by Dema · built-in
New Customer Cohort
Segment first-time buyers by channel and compute 90-day LTV
Added by Lisa E. · 1w ago
Creative Fatigue Scan
Identify ads with declining CTR over 14 days and suggest refresh candidates
Added by you · 3d ago
Multi-scenario procurement forecasting
The centrepiece of Tatti Lashes' skill library is a multi-layered procurement forecasting system. Starting from a base forecast built on historical API data and a proprietary tiering methodology, the agent layers on additional scenarios: CRM uplifts, new product launches with cannibalization adjustments, and collaboration revenue, to produce a comprehensive master procurement sheet with full monthly phasing and channel-level splits. Their most-used skill runs the entire scenario stack in a single call, producing multiple finance-ready output files and an executive summary.
Marketing budget planning
The team built and iterated their annual marketing budget entirely inside the agent, progressing through three versions. The final iteration incorporated MMM-informed channel rebalancing, introduced flexible cross-platform budget pools, and was pressure-tested line by line, arriving at a fully phased plan across 22 budget lines and 6 sections.
Automated weekly and monthly reporting
Rather than rebuilding reports manually, the team encoded their exact reporting methodology into repeatable skills. A weekly paid social report covering spend pacing, audience analysis, CPA and frequency tracking, and creative performance, runs every Monday. A full year-over-year market performance report covering revenue, category shifts, and trend deep-dives has been run 13 times. A monthly business update delivers YoY performance reporting with honest, data-driven editorial framing. And a TikTok Shop report combines platform exports with first-party Shopify data for enriched channel insights.
Product and cost intelligence
Supporting the core planning and reporting skills, the team built foundational reference skills that encode domain knowledge the agent can draw on. Bundle decomposition mapping enables accurate component-level analysis where Shopify's native data falls short. Product-to-finance category mapping replaces unreliable platform categorisation with the business's own taxonomy. And a fulfilment cost reference covering all carrier rates and 3PL pricing enables profitability analysis by market.
“We built 20 skills with the Dema agent in three weeks without manually writing a single line of code. We just described what we needed, procurement scenarios, marketing budgets, weekly reports, and iterated until the output matched. What took two people three to four months now runs end-to-end in one command, and any team member can trigger it.”
Tom Everson · Digital Director, Tatti Lashes
The Impact
In three weeks, Tatti Lashes built 20 custom skills and turned their commerce AI agent into the system that runs their planning cycle, compressing what was previously a 3–4 month process into weeks, and making analyses routine that the team had previously abandoned as too manual.
Planning speed
The procurement forecast, which previously took 3–4 months and often wasn't completed until well into the new financial year, now runs end-to-end in a single command. Built by two people, the full multi-scenario forecast was stress-tested and finalised in three weeks.
Decision quality
The scenario modelling changed real buying decisions. The team invested more heavily, and more cautiously, at the category level, moving past best-sellers to identify SKUs picking up momentum. The launch-curve modelling enabled confident procurement bets on products with as little as one month of sales data, learning from 24 months of historical launches.
Operational efficiency
The weekly paid social report now takes 10 minutes instead of an hour. Analyses the team had previously given up on entirely, like LTV mapping, cohort analysis, and comparing product launches and bundles against historic bundle performance, are now run routinely.
Knowledge retention
Domain expertise like bundle mappings, category hierarchies, carrier rates, and reporting methodologies is now encoded in reusable skills rather than scattered across spreadsheets and team members' heads.
Tatti Lashes continues to build on their skill library as the new financial year unfolds. The team is already seeing assumptions challenged and can now adjust the full forecast immediately, a feedback loop that would have taken weeks to act on before.
Ready to scale profitably?