HG Insights today put its Contextual Intelligence Platform on sale, bundling a B2B company database, sales copilots, task agents and a server through which outside AI systems can query the data. The Santa Barbara company says the release closes a gap that leaves autonomous AI agents acting confidently on fragmented, isolated records.

In Short

HG Insights, a company that sells data on which businesses use which technology, today released a bundled platform that connects that data to AI sales tools. The aim is to stop AI assistants from making confident mistakes, such as counting one company twice or treating an existing customer as a fresh prospect, which matters to sales and marketing teams that are handing more work to software. What changes is that the data, scoring that can be inspected and a connection for outside AI tools arrive in one package, sold on a consumption basis with only an entry-level figure published.

What HG Insights put on sale

HG Insights today gathered five components under one name, the Contextual Intelligence Platform. According to HG Insights, the platform consists of Fabric, the data and intelligence foundation; HG Copilots, which include Market Analyzer, Data Studio and Sales Copilot; HG Agents, which automate go-to-market tasks; the HG MCP Server; and Customer Voice, a layer powered by TrustRadius that the company describes as previously announced. Fabric, the Copilots, the Agents and the server became available today, the company said, and a refreshed brand identity debuts with them.

Rohini Kasturi, chief executive of HG Insights, framed the pitch around parity. "Every GTM team is adding the same agents, the same tools, the same playbooks. When the infrastructure is equal, whoever knows the account best wins," she said, according to the company. Her next sentence turned to the market HG Insights itself sells into: "Most vendors still sell disconnected data and call it intelligence."

The pieces have been arriving for some time. According to Business Wire, HG Insights announced its acquisition of TrustRadius, a review and buyer-intent business, in June 2025, and put out its first AI copilot the following month. PPC Land's earlier coverage shows the company's data travelling a different route: Bombora named HG Insights as one of four initial partners in its curated ecosystem audiences on October 30, 2025, with HG supplying technology adoption, spend and purchase-intent signals, and those segments later reached Reddit. Years before, under its former HG Data name, the company had pushed technographic audiences into advertising platforms through LiveRamp. The new platform points the same data at software that acts rather than at media that targets.

The context problem, as the company frames it

Much of the software now being added to sales and marketing teams is agentic AI, and the release argues that the quality of its input now separates one vendor's agents from another's. According to HG Insights, most such agents draw on isolated, fragmented internal and commodity datasets that lack entity resolution. The company lists the consequences with some precision. The same company can appear as several accounts. A subsidiary's spend may be missed, or double-counted under the wrong parent. An intent signal from an account that already owns a product can be flagged as new business instead of expansion, while one from a competitor's account can be chased as winnable. Marketing and sales can end up relying on job titles rather than evidence of who owns the product. "Without that context, an agent returns a fast, confident answer that is fundamentally wrong," the release says, adding that as more work is automated those errors scale.

An analyst quoted in the release, Michael Levy of GZ Consulting, made a related point about fragmentation: "Account data still lives across a dozen separate systems inside most organizations." The release does not say what the remark is based on.

The thesis is not unique to HG Insights. A survey covered by PPC Land found that B2B sales teams estimate 32% of their CRM data is flawed, a finding published by Firmable, itself a seller of prospect data. Another vendor, Intentsify, argued in a May 2026 white paper that agent failure traces to poor input data rather than weak models, citing a Gartner forecast that more than 40% of agentic AI projects will be cancelled by the end of 2027. Several data suppliers, in other words, are making the same argument at the same moment, and each has a product that would address it.

Fabric: the data layer

Coverage figures

Fabric is the part that can be licensed on its own, delivered by API or Direct Feed into a customer's data warehouse, CRM, applications or models, according to HG Insights. A diagram supplied with the release gives the dimensions of the data in six blocks:

  • Accounts and hierarchies: more than 70 million companies.
  • Tech footprint: more than 640 million verified installs, with 15 or more years of change history.
  • IT, cloud and AI spend: more than 150 categories.
  • Contacts: more than 200 million records across more than 190 countries.
  • Buyer intent: more than 20,000 research topics and 36,000 products tracked.
  • Tech usage by function: tracking by role and department across more than 10 million companies.

These are the company's own figures, and the material offers no audit or independent verification of them. The release adds that HG Insights' data grounds go-to-market decisions "that drive over $1.5 trillion in revenue across Fortune 500 Technology companies", a claim about influence rather than coverage that is likewise unverified.

How records move through the pipeline

The diagram describes a five-stage loop labelled data collection, curation and continuous refresh. Collect ingests global signals. Connect links technologies to accounts. Validate cross-references the records and scores confidence. Confirmcombines automatic approval with expert review. Deliver passes on what the company calls trusted intelligence. A separate input marks where a customer's first-party data enters, and TrustRadius supplies downstream buyer intent and verified product reviews. The validation and confirmation stages are where the entity-resolution claims would be tested, although the material does not publish match rates, confidence thresholds or the proportion of records that pass through expert review.

What was added

According to HG Insights, Fabric now extends its company, technology and spend coverage and adds funding and M&A intelligence, automatic hierarchy updates, and contacts matched to the buying centres that own a given technology. The last item addresses the job-title problem named earlier: a contact is tied to a group that holds a product rather than inferred from a label.

The Momentum signal

The most conceptual addition is Momentum, a signal showing whether a vendor, a portfolio or a product is gaining or losing ground. The company's description is explicit about how it reads. A declining footprint, spend and intent can flag a competitive displacement opportunity, while growth across all three shows a competitor gaining ground. The diagram groups the output into three bands: vendor, product and category momentum; multidimensional buyer intent; and expansion, retention and displacement signals. What the release does not give is a formula, a refresh frequency or the weighting between footprint, spend and intent. A single composite read is easy to consume, and equally hard to audit from outside.

Scoring that can be adjusted

HG Insights says its Fit, Need and Intent models now update dynamically while staying "deterministic, explainable, and adjustable", in contrast to black-box scoring. The two screenshots supplied with the release show what that looks like in practice, though neither carries a date range, sample size or customer name.

The first shows a segment-threshold panel. Minimum points for the very high, high and medium segments are set at 5.95, 3.67 and 2.49, with buttons labelled "Save & Compute performance" and "Auto-select the best parameters". A chart titled model performance compares each segment's share of leads with its share of conversions:

SegmentShare of leadsShare of conversionsConversion rate
Very high10%40%23.23%
High24%33%7.88%
Medium22%16%4%
Low44%12%1.55%

Read at face value, the top segment converts at roughly 15 times the rate of the bottom one, and the two upper segments together hold 34% of leads and 73% of conversions. Arithmetic on the displayed shares implies an overall conversion rate near 5.8%, a figure the screenshot does not show. The conversion shares sum to 101%, and the chart's vertical axis runs to 101, consistent with rounding. Whether the data is a customer's, a demonstration set or a mock-up is not stated, so the percentages illustrate the interface rather than evidence of performance.

The second screenshot shows the Sales Copilot view of a single account, Workday Inc, within a "No CRM Workbook". The record lists 17.7K employees, a United States headquarters, a founding year of 2005 and an estimated IT budget of $1.0B, alongside three scores labelled HF, HC and CV, each graded B (84), B (86) and B (84). The material does not define those labels. A contacts tab shows 109 contacts across 11 pages with six filters applied, listing titles such as Account Director and Chief Revenue Officer, with email addresses and phone numbers masked. The release describes the same view in prose: a rep opens a ranked account list with research, engagement history, intent signals, a suggested sales play, product pitch and buyer-committee contacts, all with stated reasoning. Sales Copilot now tailors its guidance to each customer's own products, buyers and sales motions, according to the company.

Agents, the Superagent and the MCP server

HG Agents handle tasks such as expansion sizing, territory coverage, account research, signal monitoring, brief and outreach drafting, and enrichment. They can be reached from Slack, Microsoft Teams, the HG interface, or embedded in a customer's own application using the Model Context Protocol, according to HG Insights. A new HGSuperagent interprets a request, selects specialist agents and assembles a result in which every claim carries a source. The company says an account research brief can return more than 30 data points within 90 seconds, and that a team with time for only its priority accounts can now cover an entire territory. Both are vendor statements without published test conditions. The architecture diagram describes the agents' output as grounded in HG's data, which places the quality burden on the retrieval layer beneath them.

The HG MCP Server is the component with the widest reach. According to HG Insights, it gives customers' AI agents and applications direct access to the intelligence without first building a separate data infrastructure. Through one secure, governed connection, they can query market, account and buyer intelligence alongside TrustRadius product reviews, SEC filings, federal contract data and other live web research. They can also use a growing library of curated workflows, trigger and monitor HG Agents, and query Fabric for custom analysis, market sizing, ICP segmentation, account-based marketing and campaign design. The company says every result traces back to its source data.

The release places the server inside several AI ecosystems at once. It brings the intelligence into Amazon Quick, Anthropic, OpenAI and Microsoft, so that, in the company's example, an analyst working in Microsoft, an engineering team building with OpenAI and a rep inside an HG Copilot all draw on the same picture of markets, accounts and buyers.

Where this sits in the MCP wave

Advertising has adopted MCP quickly, mostly to put campaign controls within reach of agents. Amazon opened its advertising APIs to AI agents through a server built on the protocol, Meta opened its ads server to any developer holding a Meta app on July 16, 2026, and Adform said AI assistants can reach 800 of its platform capabilities. Data owners are following. AudienceMix opened its segment catalogue to AI software through an MCP server on September 16, 2026. HG Insights belongs to this second group, in which the server exposes information rather than controls on a live account.

The protocol is also a moving target. A specification revision dated July 28, 2026 removed protocol-level sessions, and ad tech server operators had to rebuild deployments as a result. The release does not say which revision HG's server implements, how access is authenticated, or what "governed" means in terms of permissions and logging.

Customer Voice and AI search visibility

Customer Voice, powered by TrustRadius, feeds Fabric with downstream buyer-intent signals from a community of more than 12 million technology buyers who are actively researching products, according to HG Insights. It also speeds up the collection of customer-verified reviews and ratings, which the company positions as useful for social proof, SEO and AI citeability, and adds GEO monitoring to track how vendors show up across AI search.

The commercial logic is visible in recent PPC Land coverage. TrustRadius's 2026 B2B Buying Disconnect Report, which surveyed 1,862 buyers, found that 83% of technology buyers shortlisted three or fewer products, averaging 2.7, as covered in a report on AI shortlists that omit UK software firms. The same article cites a Semrush survey of 519 B2B buyers, published by PPC Land on July 20, 2026, in which 92% said AI had shaped their vendor shortlist. A review platform that also owns an account database therefore sits on both sides of the shortlist: it can show vendors where they appear in AI answers and tell sellers which accounts are researching them.

Ecosystem, availability and pricing

Fabric, HG Copilots and the HG MCP Server are available through AWS Marketplace, which, according to HG Insights, simplifies procurement and lets eligible customers apply existing cloud commitments toward the company. HG Agents and Customer Voice are not listed among the Marketplace offerings in the release, and no reason is given.

The platform is priced on a consumption model built around Intelligence Credits, which the company says gives customers flexibility, scalability and control. The one published figure is for the Copilots: the first tier of modules begins at $22,000, with larger packages sized for enterprise needs and full details available through HG Insights or its website. The release does not state whether that amount is annual or per user, how many credits it includes, or how credits are consumed by queries, agent runs or API calls. In a consumption model, those unknowns decide what a given volume of agent activity costs.

Why the release matters to the marketing community

For B2B marketers, the release pulls account intelligence toward the point where decisions are made, rather than where audiences are bought. HG's technographic data entered advertising years ago through audience segments distributed via LiveRamp. The platform instead puts the same data, plus spend, intent and contacts, behind agents that draft outreach, size territories and prioritise accounts, so that marketing and sales read from one account view.

Three questions follow from the material. First, entity resolution is the crux of the pitch, yet it is the area where the release publishes the least: no match rates, error rates or benchmarks against rival datasets appear. Who verifies the match, then? Second, explainable scoring is promised without a stated method for validating the models, and the supplied screenshots lack the sample details needed to judge them. Third, a server reachable from several AI platforms lowers one barrier, since customers need not standardise on a single AI stack, but it makes the data's pricing and permissions the real point of lock-in, and those are the details left open.

None of this is a finding about the product's performance, which no independent party has yet tested in the material available. It is a description of what a data supplier now says it will offer to software acting on a buyer's behalf, at a time when PPC Land's coverage of MCP shows ad platforms and data owners alike opening servers for agents to query.

Timeline

  • June 2025: HG Insights acquires TrustRadius, according to Business Wire.
  • July 2025: HG Insights unveils its first AI copilot, according to Business Wire.
  • October 30, 2025: Bombora names HG Insights among four initial partners in curated ecosystem audiences.
  • May 2026: Intentsify publishes a white paper arguing that poor input data is the main failure mode of AI agents in B2B sales and marketing.
  • July 11, 2026: Adform's solutions engineers publish 29 agentic skills on GitHub for its MCP server.
  • July 16, 2026: Meta opens its ads MCP server to any developer holding a Meta app.
  • July 28, 2026: A revision of the MCP specification removes protocol-level sessions.
  • September 16, 2026: AudienceMix opens its segment catalogue to AI software through an MCP server.
  • Today, October 6, 2026: HG Insights releases the Contextual Intelligence Platform, with the announcement scheduled for 10am PT (1pm ET), according to HG Insights.

Summary

Who: HG Insights, a Santa Barbara-based provider of B2B technology, spend and buyer-intent data led by chief executive Rohini Kasturi, with TrustRadius supplying review and buyer-intent data and with AWS Marketplace, Amazon Quick, Anthropic, OpenAI and Microsoft named as channels or ecosystems.

What: The Contextual Intelligence Platform, comprising Fabric, HG Copilots, HG Agents, an HG MCP Server and the TrustRadius-powered Customer Voice, with a new Momentum signal, deterministic and adjustable Fit, Need and Intent scoring, and an HGSuperagent that routes requests to specialist agents.

When: Today, October 6, 2026, when the platform became available, according to HG Insights.

Where: Through direct API or Direct Feed delivery, the HG interface, Slack, Microsoft Teams, AWS Marketplace and customers' own AI applications via the MCP server.

Why: HG Insights says AI agents in go-to-market work act on fragmented data that lacks entity resolution, producing fast but wrong answers that scale as automation grows. The release does not publish independent evidence of accuracy, and several pricing and technical details remain open.