We build the AI solutions clients actually use

Outcomes, not pilots

From consumers to enterprise buyers, countless decisions now start inside ChatGPT, Gemini, Claude, Copilot, Google AI Overviews and AI Mode, often before they reach a website. For most brands, website visits from traditional search have fallen by as much as 50%, and communications teams have no way of knowing whether they appear in those answers, in what context, or alongside which competitors. The same squeeze applies either side of that moment. The account research that should inform a pitch or a campaign takes weeks of manual work, so it arrives late or not at all. Audience research is rationed by cost and time, so the questions that come up daily, does this headline land, will the procurement lead care about this proof point, never reach an audience. They get answered by the most senior person in the room.

The AI Lab is Hotwire’s product and engineering team: AI engineers, data scientists, developers, and communications and marketing strategists working in one group. We build the software ourselves rather than reselling a vendor’s platform, because the questions communications teams ask are not the questions generic marketing tools were designed to answer. Every product starts from a live client problem and is judged on whether it changes a decision.

Our proprietary tech stack is designed to answer real questions: visibility, optimisation, deep intelligence and real time testing. Spark monitors brand visibility across AI answer engines and returns the topics, media and analyst voices that shift it. Radiate rebuilds content so those engines cite it, scoring each page and returning an optimised version with schema markup. Ignite runs multiple AI agents across deep web sources and internal systems including CRM, intent signals and technographics, compressing market research, competitive analysis, SWOT assessments and persona development from weeks into days. Ember builds synthetic audience personas from real evidence and lets teams question them directly. Equinix, EY, KPMG, Telstra International, Cloudflare, Flexential, Everpure and Continental use Spark. ROI·DNA sells Spark, Radiate and Ignite to its own clients under the ROI·DNA name.

What we built, and what it fixes

Marketing and communications teams cannot act on what they cannot see. Spark monitors how a brand and its products appear across ChatGPT, Gemini, Claude, Copilot, Google AI Overviews and AI Mode, tracking mentions and citations, the context around them, which competitors surface alongside, and which sources the engines cite. It returns recommendations teams can act on, plus the media and analyst voices that most influence answers in their category. Spark runs as a managed service.

Content written for search engines is not structured the way answer engines read it. Radiate scores a page across four areas: structure and clarity, quotability and tone fidelity, FAQ quality and coverage, and metadata signals. It returns a citability score out of 100 and an optimised version of the content with the schema markup code to deploy. Radiate works on published URLs and unpublished drafts of 500 to 2,000 words.

A pitch or a campaign plan is only as good as the account research behind it, and that research usually takes weeks nobody has. Ignite deploys multiple AI agents across deep web sources and internal systems, including media intelligence, CRM records, intent signals, technographics and competitive intelligence. This isn’t the off-the-shelf Claude or ChatGPT. It returns market research, competitive analysis, SWOT assessments, persona development and deep intelligence in a fraction of the time the manual equivalent takes.

Audience research is rationed by cost and time, so most communications decisions get made without it. Ember builds synthetic personas from first-party research where a brand has it, and from AI-augmented research across public sources where it does not. Teams question them directly: test a concept, convene a buying group as a panel, and ask any persona why it answered as it did. It complements primary research rather than replacing it.