AI Sales Agents Fail Without Context: How to Ground GTM Agents With HG MCP Server

AI sales agents can produce a polished account plan in seconds, and that is exactly the risk when the facts underneath are wrong. At Context26, HG Insights VP of AI Strategy Francis Brero showed how a confident, well-formatted brief can still send a seller toward the wrong strategy, and how grounding agents in HG Insights Contextual Intelligence through HG MCP Server fixes the inputs before the reasoning starts. This recap covers his session.

Quick Answer: AI sales agents give confident wrong answers when they reason over incomplete or incorrect account data. Grounding them in a governed context layer, such as the Fabric exposed through HG MCP Server, gives them accurate installs, spend, and account facts before they reason, so the strategy they recommend fits the account.

Key takeaways

• Wrong context is not neutral. According to the HubSpot numbers he cited, an agent with wrong context performs worse than no AI at all.
• Errors hide in polish. Small mistakes, such as calling an expansion account net new, lead the agent to reason toward the wrong strategy.
• Better models do not fix bad inputs. Reasoning applied to incorrect facts still leads a team astray.
• Two access modes fit two jobs. Purpose-built MCP Tools answer account questions, and Fabric queries answer market questions.
• Starting is free. More than 100 GTM use cases are available as prompts that teams can modify and run in their own ecosystem.

What Francis shared at Context26

Francis followed the CTO’s session by turning to the practical question: how do you put HG Insights context inside the agents your teams are already building? He introduced HG MCP Server and the HG Agents library, explained why wrong context does damage that no model can undo, and showed how teams can start building with both. To see the full session, watch the Context26 recording.

Why wrong context is worse than no AI for GTM agents

Anyone who has used a large language model (LLM) knows it can produce great-sounding answers that are not true. Francis pointed to numbers HubSpot shared a few weeks earlier at its INBOUND event, which he summarized in one line.

“You’re better off not using AI than using AI with wrong context.” – Francis Brero, VP of AI Strategy, HG Insights

The point, in his framing, is that wrong context is not a neutral gap. It is detrimental to the success of any initiative that uses AI. That is why HG Insights built HG MCP Server: to bring context that grounds the agents teams build on the go-to-market side, so that when a seller goes after an account, the competitive landscape and the incumbents are clear.

What a wrong account brief looks like (and why the errors are hard to spot)

Francis walked through an anonymized scenario he sees often. A seller at a major cloud provider asks an LLM for an account plan on a large enterprise account. The result is a beautiful asset with a convincing story, and the polish makes it easy to believe.

“It’s not obviously wrong in a way that you can catch immediately, which actually makes it that much more dangerous because the errors are small, but they can lead to the wrong kind of reasoning.” – Francis Brero, VP of AI Strategy, HG Insights

The team compared the brief against HG Insights data and found two mistakes. The LLM treated the account as net new, when the account had already been a customer for years, which makes it an expansion. It also named the wrong main data platform. Each error looked minor in isolation.

The cost shows up downstream. The brief’s strategy was built around landing a new customer, when the account called for expanding an existing relationship, and Francis said there was a much better story to tell once the right context was in place.

“If you feed them the wrong context, their incredible reasoning is being applied on incorrect facts.” – Francis Brero, VP of AI Strategy, HG Insights

Models are getting better at reasoning, he said, and that is the trap. Strong reasoning on wrong facts leads a team astray, destroys trust with sellers, and damages trust with the customer, because the seller walks into the meeting with the wrong assumptions.

What a GTM-grade MCP server gives an agent

A Model Context Protocol (MCP) server is how an agent reaches outside data. HG MCP Server gives agents a single connection to the entire Fabric, Francis said, with all of its depth and breadth across companies.

Through it, an agent can reach company data, AI maturity, Cloud Spend, Tech Installs, and historical IT Spend. It can also read specific sections of SEC filings that point to initiatives worth pursuing. In practice, a seller preparing for an account can ask what the competitive landscape looks like, who the incumbents are, and how long each technology has been installed.

Francis tied this to Functional Area Intelligence (FAI), which Satish covered earlier in the day. FAI shows where decisions are made and who holds the power, and those are the data points that let agents help with strategic account targeting, not just research.

Purpose-built tools versus asking questions of the whole Fabric

HG MCP Server changes how people interact with HG Insights. Historically, Francis said, HG Insights provided data points about specific accounts, and that is still available through purpose-built MCP Tools. A seller researching one account’s ERP landscape can call the Tech Installs tool for that account and get the answer directly.

The second mode is querying the Fabric at large. Instead of looking up one account, an agent can answer market-level questions: how the CRM landscape is evolving in mid-market healthcare, which vendors are gaining momentum, and which are losing it. Francis said teams can run this kind of market analysis directly through agents.

The two modes serve different jobs. A purpose-built tool fits a question about one account, and a Fabric query fits a question about a market. Both draw on the same Fabric. His broader point was that the quality of the context decides the quality of the result.

“There’s no amount of compute that can compensate for bad context.” – Francis Brero, VP of AI Strategy, HG Insights

100+ free GTM agent prompts you can download and modify

Through years of working with enterprises, Francis said, HG Insights has built a clear picture of the jobs to be done in go-to-market. HG Agents answers those jobs out of the box, including an account brief for a seller and an ICP, so a team gets to value quickly.

At Context26, HG Insights announced it is making the harnesses and prompts behind these agents available on its new website. According to Francis, that is a little over 100 GTM use cases, spanning three kinds of work:

• Account workflows: preparing for a pre-meeting, and researching an account based on its competitive landscape or on big initiatives announced in SEC filings.
• Product marketing: building competitive battle cards using Customer Voice, powered by TrustRadius, review data to show how a product compares for a specific segment.
• RevOps: improving ICP scoring and territory planning.

The prompts are free to access, and signing up for HG MCP Server is also free and includes starting credits. Francis wanted teams to see everything.

“I’m really excited to make this kind of open source where there’s nothing hidden. All of the prompt is available so you can use it.” – Francis Brero, VP of AI Strategy, HG Insights

Teams can download the prompts, modify them, and run them in their own AI ecosystem as the starting point for their own agents.

Where HG MCP Server and HG Agents run: Anthropic, OpenAI, Microsoft, Amazon, and your own

HG MCP Server connects to the ecosystems where teams build agents. Francis named the frontier labs, including Anthropic and OpenAI, and the major platforms from Microsoft and Amazon. It also connects to agents companies build internally, which he said many HG Insights customers already do so their agents always have the right context for their decisions. In partner ecosystems, it appears as HG Insights Connect, powered by HG MCP Server.

In response to an audience question about Microsoft, Francis described the benefit as a single connection. Whichever ecosystem a team works in, it reaches the same Fabric, so the intelligence behind each agent stays consistent.

Watch the full session

This recap covers Francis’s session. Watch the full Context26 recording to see the examples in context. To see how HG Insights supports agent builds, explore GTM infrastructure for AI agents or visit the Contextual Intelligence Platform page. To start building, download the free agent prompt library and sign up for HG MCP Server free credits