Unlocking the Web: Research Agents

Agentic has been the biggest technology buzzword of the year. The ability to automate routine admin tasks is the holy grail of enterprise automation, because it frees up human workers to focus on high value activities. The real world is not so clear cut. Most marketing teams have spent the last twelve months struggling with the age old problems of data quality and target account selection. Two different worlds that will eventually converge when Agentic AI is production ready.

Except that AI can help data cleansing right now, and has done for many years. Predictive normalisation is a well established technique that can populate individual profile fields automatically. It has assisted with job level and job function calculation for over a decade. Elsewhere, intent data has become a key targeting criteria for many ABM programs. Here too, predictive AI is used to interpret the vast wealth of intent signals in order to discover meaningful trends and enable account-level insights.

AI Search Engines

This year, another enterprise ready use case for AI has emerged that once again promises to fill the profile gaps that traditional data solutions can't reach. Web search has been the generative AI success story of 2025. Google's declining market share is the hot topic dominating marketing discussions. Meanwhile, ChatGPT can generate detailed podcasts or multi-page research reports containing everything that a sales rep needs to prepare for a meeting, all sourced from web articles, social media conversations and increasingly from internal documents.

That technology is now being integrated into the martech stack. Over the summer, HubSpot announced a direct integration between their platform and ChatGPT Deep Research. That allows SDRs and sales reps to generate research reports for any account directly from CRM. Marketers can use this capability too. A new breed of GTM data platforms is making this possible. AI doesn’t just segment your database. It can enrich it with web information, as well as from external data providers.

Account Data is Everywhere

Everything you need to know has been published online. ChatGPT has a wealth of information about every household name enterprise. Now that data is available for marketers to use. With a research agent, it's far easier to get niche firmographic data from the web than it is from D&B. Most leading brands will have web pages detailing their contact centre strategies to customers and customer counts to investors. Meanwhile, top executives will have published social posts discussing suppliers and customer relationships. All data that is invisible to traditional data brokers, but easily accessible to AI search engines.

Research Agents also help address another common issue with purchased data: recency. ChatGPT has direct access to the most up-to-date information about startups and small enterprises not typically found in the leading data providers. Such businesses will have a web presence, but won’t appear in most firmographic databases. For the right product, they're a greenfield market invisible to traditional data management approaches. That’s where research agents can help. They use AI search engines to scrape Linkedin and company websites in real-time, which in turn delivers better match rates for common firmographic profile fields such as industry or employee size, particularly in niche verticals.

Use Multiple Solutions

Finding the right information about midsize businesses in small markets has always been a challenge. The leading data sources typically only have limited coverage, and match rates are poor. Some European markets do have local data providers available, but that’s not true everywhere. Using AI as a secondary data provider helps close those gaps, enriching more of your database than is possible through traditional methods. Such techniques introduce the risk of hallucinations, but then the best data brokers are often wrong too. That’s not stopped marketers from using purchased account lists.

Ultimately, good data drives modern marketing. Better data leads to deeper personalisation and better outcomes. Yet, actually improving data quality is an ongoing challenge. The best data improvement programs rely on multiple sources to fill the gaps. That’s particularly true for marketers in small markets or niche industries, who have long struggled to uncover their target account lists. Help is at hand. Large language models have opened up an alternative avenue to solving data completeness challenges. Generative AI is not a magic bullet, and won’t have complete coverage, but it can be used to supplement your existing data strategy.