Hightouch, a seven-year-old startup focused on marketing production, has reached $100 million in annualized recurring revenue (ARR) after adding $70 million in ARR since launching an AI-powered content creation service in late 2024. The company says its AI tools allow marketing professionals to create custom images and videos for ad campaigns without involving brand design teams or ad agencies. According to TechCrunch’s interview with co-CEO Kashish Gupta, the platform connects directly to a brand’s existing creative workflow to ensure brand-safe advertising.
From design teams to AI-assisted marketing production
Historically, marketers relied on designers and other creative professionals to develop images and videos for personalized online ad campaigns. Hightouch’s model shifts that workflow by offering an AI service that creates custom content for brands including Domino’s, Chime, PetSmart, and Spotify without involving brand design teams or ad agencies.
In late 2024, Hightouch launched this AI-powered service. Since introducing its AI product 20 months ago, the company added $70 million in annualized recurring revenue, bringing total ARR to $100 million. The company’s leadership includes co-CEO Kashish Gupta and co-CEO Tejas Manohar, who previously served as an engineering manager at Segment—a customer data platform acquired by Twilio for $3.2 billion in 2020.
From a technology standpoint, Hightouch addresses a usability gap. Gupta told TechCrunch that before GenAI, it was “impossible for someone without many, many years of design skills to create consumer-level assets.” This points to a common industry challenge—turning AI output into assets that meet creative standards—while positioning Hightouch’s software as a bridge between marketing teams and production-grade design.
Why foundation models fall short for brand-safe ads
A central technical argument is that Hightouch’s approach goes beyond what standard AI models can do on their own. The company says brands initially attempted to generate ad campaigns using general foundational models—broad systems that power tools like chatbots but lack knowledge of specific brands—only to find the resulting images and videos failed to meet “on-brand” standards.
According to Gupta, those attempts produced two types of failures. First, foundation models “didn’t know about specific consumer brands,” including “colors or fonts, tone, or assets.” Second, he says “LLMs would hallucinate products that didn’t exist,” and this breaks practical requirements for advertising and email marketing: “you can’t do advertising and emails on products that don’t exist.”
For technology teams and marketers, the implications are straightforward: when output must match a brand’s identity system and product catalog, a model that improvises content without grounding can create assets that are unusable. Hightouch’s design goal is not only generating images and videos, but constraining generation so that results align with known brand materials and real product references.
Direct connections to existing creative tools
To achieve brand consistency, Hightouch connects directly to customers’ existing creative tools. The list includes Figma, photo libraries, and content management systems (CMS). According to the company, pulling from these sources allows the platform to “learn” a company’s specific brand identity.
From there, Hightouch’s AI agents use photos, designs, and customer insights to help marketers build personalized ad campaigns. The reported outcome is a workflow where marketing teams do not have to wait on designers or developers, while the system aims to keep assets consistent with the brand’s established look.
Hightouch illustrates this approach with a concrete example. Gupta told TechCrunch that “Domino’s will never generate a pizza.” Instead, he says the company “will always use existing images of pizza,” then place that image into an ad where the background and other surrounding elements might be generated. In technical terms, the workflow treats certain asset classes—like product imagery—as fixed inputs, while leaving parts of the composition to generative steps.
This design choice addresses both brand identity and factual consistency. By drawing product visuals from known libraries, the system reduces the risk that the generator invents nonexistent items—an issue Gupta attributes to hallucinations in foundation-model-only workflows.
Funding and market traction
Hightouch’s reported traction includes its headcount and funding context. The company employs approximately 380 people, and was valued at $1.2 billion in February 2025 when it raised an $80 million Series C round led by Sapphire Ventures.
These figures suggest that customers are paying for a specific production workflow rather than using generic generative tools. The underlying technical approach—connecting generation to existing creative systems—indicates a practical direction for the category: AI that functions as a production assistant constrained by brand inputs, rather than a standalone generator operating without context.
For marketers, this represents a shift in how creative work is operationalized: instead of transferring tasks to specialized designers, marketing teams can request personalized creative assembled from brand-controlled components. For developers building in this space, the integration surfaces (Figma, photo libraries, CMS) appear central to making generative output usable in real campaigns.
Source: TechCrunch