Building CreativeX: AI-powered brand governance at scale

Executive overview

Brands now produce hundreds of thousands of creative assets per year across fragmented digital platforms. Human oversight alone can no longer catch brand inconsistencies, quality failures, or representation issues at that volume. CreativeX uses computer vision, OCR, and audio parsing to give brands structured data on every piece of content they produce.

The core insight: creative governance — not performance optimisation — is the unsolved problem for large brands.

How CreativeX works

• Combines computer vision, OCR, and audio parsing to extract structured data from images and video
• Builds brand-specific algorithms to detect consistency, quality, and representation issues
• Delivers a data layer across every creative execution, not just selected campaigns
• Clients include Unilever, Pepsi, Heineken, Mars, Nestlé, Renault, Facebook, and Google

The origin story: pivot from a failing e-commerce business

• Anastasia left Google in 2015 to build Hatch, a customisable e-commerce marketplace
• The business was running out of money; she was too embarrassed to admit failure
• The team built internal tools to understand why certain creatives converted better
• That experiment salvaged the unit economics of Hatch
• Investors at a final-chance pitch meeting spotted the tool and said: "That's your business"
• A term sheet arrived two days later — no company name, no pitch deck

Finding the real product through customer feedback

• Early versions of CreativeX focused on performance optimisation (clicks, conversions)
• Customers kept buying ad hoc but refused annual contracts
• Direct question to customers surfaced the real problem: brand control during content scale
• 12 product versions before product-market fit on brand governance
• Shift from "drive more clicks" to "help me control the flood first"

Mistakes and hard lessons

• Left Google on a Friday and incorporated a new company on Saturday — no break, seven years without time off; take a break before founding
• Chose investors based on VC firm prestige rather than individual alignment
• Prominent VCs expected a straight-line growth path; when it didn't materialise, they disengaged
• Lesson: optimise entirely for the individual partner, not the brand name of the firm
• Company values were dismissed early as "posters on a wall"; rapid growth from 10 to 50 people proved their necessity
• Values gave a shared vocabulary to address performance and culture issues without personal blame

Representation in advertising: unintended consequences at scale

• After Black Lives Matter and MeToo, brands increased representation of women and people of colour in ads
• Simultaneously, stereotypes increased — women shown in kitchens, people of colour shown only in athletic contexts
• Good intentions do not guarantee good execution; measurement is required to verify the outcome
• Research identifies a marketer empathy gap: marketers assume their audience mirrors themselves, but household income, education, and values often differ significantly

Scaling leadership and the CEO transition

• Skills that take a company from 0 to 50 people are not the same skills needed to go from 50 to 500
• Anastasia describes herself as an "underconfident overachiever" — identity tied to doing, not directing
• The transition requires shifting from execution to enabling others, which feels threatening to high-output founders
• Four-discipline leadership flywheel: inspiring → engaging/selling → planning → coaching
• Headline exercise: desired future headline was "best place to work", not a revenue milestone
• Implemented a four-day work week as a deliberate bet that happy, balanced people produce better work

Culture and company values

• CreativeX values data-driven thinking and independent dissent over harmony-seeking
• Values provide a filter: attract people who want to debate; repel those who don't
• Transparent values allow candidates to self-select — "take it or leave it, this is who we are"
• Shared vocabulary from values enables honest performance conversations