Typewise CEO says AI audits put the company in only 3 of 110 search results

Typewise CEO David Eberle says an internal audit of frontier large language models found the company appeared in just 3 of 110 customer-service infrastructure recommendations, far behind Zendesk and Intercom. The findings underscore how AI systems can steer enterprises toward legacy tools as autonomous service agents become more common.

Why it matters: - Typewise says the audit shows a structural visibility problem in enterprise software discovery, not just a marketing gap. - The results suggest large language models may keep steering companies toward legacy customer-service suites even as autonomous AI agents become the users and operators of those systems. - The mismatch matters because AI-driven service workflows rely on infrastructure recommendations that can shape procurement and deployment decisions.

What happened: - Typewise CEO David Eberle shared results from an internal audit of how frontier large language models recommend enterprise customer-service infrastructure. - The audit tested 110 standard customer-service queries across GPT-5.4 mini, Claude Sonnet 4.6, Gemini 3.5 Flash, Grok 4.3 and DeepSeek V4 Flash. - Typewise appeared in 3 responses. - Zendesk appeared in 85 responses. - Intercom appeared in 82 responses.

The details: - Eberle said the models were trained on legacy platform documentation and learned to recommend suites that enterprises have historically bought. - The audit found that some models recommended product lines that have already been phased out. - Typewise describes its platform as a full AI agent system that runs across existing company systems and resolves customer requests end to end, with human approval where needed. - Typewise says its system connects to CRM, billing and commerce tools already in use. - The company says an AI supervisor coordinates specialist agents that handle support, sales and commerce across email, chat, phone and messaging. - Typewise says the 3-of-110 result reflects outdated training data rather than weak visibility in the market. - Typewise notes that autonomous shopping agents, billing-dispute handlers and customer-service proxies are already in production.

Between the lines: - The audit points to a deeper issue for fast-moving enterprise categories: LLMs may surface the most familiar tools, not the most current ones. - If autonomous agents are the ones asking for infrastructure recommendations, outdated model outputs could create operational risk at scale. - Eberle framed the issue as a broader market problem, arguing that AI systems are still better at citing legacy platforms than identifying AI-agent-native ones. - The findings also suggest some enterprise buying decisions may be shaped by what large language models can recognize, not just what vendors can prove.

What's next: - Typewise is exploring whether to make the audit methodology available to other enterprise software vendors. - The company appears to be positioning the audit as a tool for measuring how AI systems currently rank software categories. - As more autonomous agents enter support and commerce workflows, the gap between legacy recommendations and current infrastructure needs is likely to stay under scrutiny.

The bottom line: - Typewise is arguing that its low appearance rate in LLM recommendations is evidence of a broader infrastructure discovery problem, not a lack of relevance. The bigger issue, in the company’s view, is that AI systems are still recommending yesterday’s customer-service stack for tomorrow’s workflows.