The Brewers just won their first World Series. As the Sausage Race proves, you never know who wins until the finish line. The AI race is no different. Threekit works with Grok, Claude and OpenAI - and we don’t pass the token costs on to you. #Brewers #WorldSeries

What does it cost to run a frontier model as a customer support agent? Our @G2dotcom agent evaluations team measured it. We built an agent on 12 foundation models (22 configurations) from Anthropic, OpenAI and Alibaba. Each got the same one-page support policy and 38 live tools, and ran the same tasks as the vendor products on our CX AI agent leaderboard. Then we priced every token at list API rates. Five takeaways: 1. 𝗧𝗵𝗲 𝗯𝗲𝘀𝘁 𝘀𝗰𝗼𝗿𝗲 𝗱𝗶𝗱 𝗻𝗼𝘁 𝗰𝗼𝗺𝗲 𝗳𝗿𝗼𝗺 𝘁𝗵𝗲 𝗺𝗼𝘀𝘁 𝗲𝘅𝗽𝗲𝗻𝘀𝗶𝘃𝗲 𝗺𝗼𝗱𝗲𝗹. Claude Sonnet 5.5 scored highest at 98.6% for 3.4¢ per conversation. Claude Opus 5.5 cost 2.4x as much at the same effort setting for a slightly lower score. 2. 𝗠𝗼𝗱𝗲𝗹 𝗰𝗼𝘀𝘁 𝗶𝘀 𝘀𝗺𝗮𝗹𝗹 𝗻𝗲𝘅𝘁 𝘁𝗼 𝘁𝗵𝗲 𝘁𝗶𝗰𝗸𝗲𝘁. 0.4¢ to 8.7¢ per conversation, vs. $0.99 per resolution for a leading CX agent and $5 to $10 for a human-handled ticket. 3. 𝗡𝗲𝘄𝗲𝗿 𝗺𝗼𝗱𝗲𝗹𝘀 𝘀𝗰𝗼𝗿𝗲 𝗵𝗶𝗴𝗵𝗲𝗿 𝗔𝗡𝗗 𝗰𝗼𝘀𝘁 𝗹𝗲𝘀𝘀. Across four Sonnet releases, the score rose about six points while cost per conversation fell 27% to 43%. 4. 𝗧𝘂𝗿𝗻𝗶𝗻𝗴 𝘂𝗽 𝗲𝗳𝗳𝗼𝗿𝘁 𝗺𝗼𝘃𝗲𝗱 𝗰𝗼𝘀𝘁 𝗺𝗼𝗿𝗲 𝘁𝗵𝗮𝗻 𝗾𝘂𝗮𝗹𝗶𝘁𝘆. GPT-5 mini at high effort cost 4x as much for 0.7 points. 5. 𝗧𝗵𝗲 𝗯𝗶𝗹𝗹 𝗶𝘀 𝗺𝗼𝘀𝘁𝗹𝘆 𝗶𝗻𝗽𝘂𝘁 𝘁𝗼𝗸𝗲𝗻𝘀. The policy, tool definitions and conversation history are present every turn, so cached-input pricing sets the economics. The line I find most telling: the gap between the cheapest configuration and the best is about 3¢ per conversation, and 16 points of quality. For a support desk, choose quality first. Caveat: these are Tier 1 support tasks. Longer, harder tasks will raise costs for every model, and we expect the gaps to widen. Thank you to @cyrildelattre and our G2 agent evals team for building this. Full analysis and chart: ai.g2.com/evaluations/insigh…

Most B2B companies spend $1 on winning AI answers (AEO) for every $40 on search (SEO and SEM). Our @G2dotcom Executive Advisory Board says that has to change, fast… Thank you to our EAB members for your candor in Napa last week as you pressure-tested our Reach story and age-of-AI strategy. The consensus: agentic buying is here, and winning the AI answer is now a go-to-market priority, not a marketing experiment. Here are the five moves we recommend: 1. 𝗙𝗶𝘅 𝘁𝗵𝗲 𝗽𝗿𝗼𝗺𝗽𝘁 𝗽𝗿𝗼𝗯𝗹𝗲𝗺. Track the prompts buyers actually type, not your own category language. Ask every inbound lead, "What prompt did you use?" 100 real prompts will reshape your content plan. 2. 𝗦𝗼𝗹𝘃𝗲 𝘁𝗵𝗲 𝗲𝘃𝗶𝗱𝗲𝗻𝗰𝗲 𝗽𝗿𝗼𝗯𝗹𝗲𝗺. About 85% of AI citations come from third-party sources. Specific, verified, and fresh peer evidence beats volume. 3. 𝗥𝗲𝗮𝗹𝗹𝗼𝗰𝗮𝘁𝗲 𝗯𝘂𝗱𝗴𝗲𝘁. Move your search-to-AEO split from about 98/2 today to 85/15 next year and 75/25 by 2028. 4. 𝗖𝗵𝗮𝗻𝗴𝗲 𝘁𝗵𝗲 𝗺𝗲𝘁𝗿𝗶𝗰𝘀. MQL volume won't come back, but quality and conversion will. Measure win rate, deal size, and velocity. 5. 𝗧𝗿𝗮𝗰𝗸 𝗮𝗴𝗲𝗻𝘁𝘀 𝗮𝗻𝗱 𝗵𝘂𝗺𝗮𝗻𝘀 𝘀𝗲𝗽𝗮𝗿𝗮𝘁𝗲𝗹𝘆. Agents don't respond to emotional marketing. The point I find most telling: being cited is not the same as being chosen. Being the recommended answer converts 10 to 70 times better than a mention. Boards assume AI should shrink marketing budgets. The opposite is true. This is marketing's moment in the boardroom. What's your AEO-to-search spend ratio today? We'll go deeper at G2 REACH in San Francisco on Nov 10. reach.g2.com/

The four Stoic virtues in daily life… Courage: speak up, stand up, have skin in the game Temperance: fasting, exercise, get up early Justice: vote, give, leave this place better than you found it Wisdom: always stay a student, always be reading

It’s great to see @G2dotcom, @Capterra, and @SoftwareAdvice collectively rank #1 for the most-cited sources in @promptwatch's latest analysis of software-related prompts in @ChatGPT. Together, the three G2 marketplaces account for 1.23% of citations — slightly ahead of @github at 1.20%. G2 alone leads all B2B review sites and ranks third at 0.73%. 💭 Companies used to ask: how do we rank on page one of Google? Now they also need to ask: does our product show up in answers from AI assistants like ChatGPT? ✨ AI answers draw from a mix of sources, and for software companies, showing up takes more than clear product copy on your own website. It also takes evidence from real customers. That’s what we’ve built at G2 over the years: verified reviews, comparison data, and buyer insight that help give people, and AI agents, trust signals to make better software decisions. I’m proud to see G2’s brands show up in the sources ChatGPT uses. It’s a strong signal that customer voice is becoming part of how AI builds a software shortlist. Read the full report: promptwatch.com/data/chatgpt…

Fun visit with our daughter thriving at the Farm ❤️ 🌲#gostanford

More than 1,200 users across 550+ organizations already query G2 data through our MCP integrations with tools like @ChatGPT, @claudeai, @profound, @HubSpot, @Gong_io, and more. Now, we’ve added the G2 MCP to @MSFTCopilot. For teams whose company has standardized on Copilot as its one approved AI tool, this is a real change. Software research no longer means leaving the tools in your workflow just to check a vendor, a review, or a competitor. Ask a question in plain language, and Copilot answers with @G2dotcom data with verified G2 data like reviews, ratings, and buyer behavior. 🔎 𝗙𝗼𝗿 𝘀𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝗯𝘂𝘆𝗲𝗿𝘀: Verified reviews and ratings, product, vendor, and category data, and Grid Reports to compare products side by side. Research boards let you build a shortlist and share it with your team. 🎯 𝗙𝗼𝗿 𝘀𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝘀𝗲𝗹𝗹𝗲𝗿𝘀: Buyer Intent and Market Intelligence show which products and competitors buyers are actually evaluating. Buyers are asking AI tools which software to trust, and those tools are only as good as their sources. We think the source should be real customer reviews. Copilot now gives teams that answer, with G2’s verified reviews, right where they already work. That’s how we help buyers and sellers reach their next PEAK. Check out the demo below to see how it works and ready our blog here: sell.g2.com/resources/succes…

As entrepreneurs, we spend much of our careers competing for investor love and attention. Fun to flip the script with our new @G2dotcom Investor Leaderboard: which VC and PE portfolios are winning with the people who actually buy software? Today's top 5 by Investor Score: 1. @sequoia (78) 2. @SalesforceVC (75) 3. @insightpartners (73) 4. @Accel (72) 5. @ICONIQCapital (71) Behind the board: 50 firms, 4,966 portfolio companies on G2, 969 G2 Leaders. The score averages portfolio quality (share of Grid-rated companies that are Leaders) with presence (how many Leaders). Insight has the most Leaders of any firm at 153. Sequoia's 93 Leaders are 63% of its rated portfolio. That is what puts Sequoia first. The input I find most telling is AI visibility. More buyers now start by asking ChatGPT, Claude, or Gemini which software they should buy. The typical G2 Leader is named in 41% of AI answers about its category. The typical non-Leader is named in under 1%, and nearly half were not named in a single answer in the last 30 days. If your portfolio company is not in the answer, it is not in the deal. Thank you to Ruehl Johnson, our AI builder intern in the Office of the CEO, who built the whole data pipeline with Claude: 12,000+ holdings verified against G2's catalog, blind-tested against firms with known answers, now running unattended. Half a billion tokens well spent. His build note on the method, including the three scoring models that did not hold up. Thank you to @thejamescad and the @profound team for the profound insights and partnership that power the AI answer analytics on this board. And to Fiona Erickson and @sanderssays, Ruehl's thought partners on the AEO rankings behind it. About 3,700 B2B companies in these portfolios are not yet on G2. Getting them listed and reviewed is the fastest way to move a firm up. It's a beta, so tell us what you would change. Investors: where does your firm stand? Founders: is your investor on the G2 leaderboard? See link to live leaderboard here: ai.g2.com/investor-leaderboa…

a few weeks ago we launched G2 CX agent evals in beta for purpose-built agents. we just added a model baseline: an agent on a model using our standard harness, and run the same tasks as vendor products. the hard part is picking models a support desk could afford to run at scale. we show @AnthropicAI Claude Haiku 4.5 and @OpenAI gpt-4o mini by default, and let you pick more advanced models and a range of effort for each. same setup as the agents: one-page policy as instructions, mcp access to the same 38 tools, same 46 scenarios, 4 runs each. on policy compliance, Haiku 4.5 scores 83% and gpt-4o mini 67%, vs 87 to 92% for @tidiocx, @BoldDesk, @retellai, @LiveAgent and Chipp. the baseline measures the model only. vendors ship integrations to ticket system and crms, escalation logic around the model, deal with edge cases learned from real deployments. we’re also building more complex tasks to capture that in the scores. i'm looking for product folks who have run these model vs. agent comparisons, to swap notes on approach. if you know one, tag them!

𝗔 𝗰𝗼𝗺𝗺𝗼𝗻 𝗮𝘀𝘀𝘂𝗺𝗽𝘁𝗶𝗼𝗻 𝗶𝘀 𝘁𝗵𝗮𝘁 𝗰𝗼𝗺𝗽𝗮𝗻𝗶𝗲𝘀 𝗵𝗮𝘃𝗲 𝘁𝗼 𝘁𝗿𝗮𝗱𝗲 𝗔𝗜 𝗮𝗱𝗼𝗽𝘁𝗶𝗼𝗻 𝘁𝗼 𝗲𝗳𝗳𝗲𝗰𝘁𝗶𝘃𝗲𝗹𝘆 𝗺𝗶𝘁𝗶𝗴𝗮𝘁𝗲 𝘁𝗼𝗸𝗲𝗻 𝗰𝗼𝘀𝘁𝘀. We conducted 100+ AI-led interviews with US enterprise leaders who evaluate, select or fund LLMs about how they manage LLMs and AI budgets, and what they described shows how difficult that tradeoff can feel. One senior director at an IT managed services company told us they're moving from a culture of "use AI for everything, to use AI where it makes sense," because "the bill has gone very high very quickly." Of the leaders who discussed cost per outcome, about one in four described cost-driven limits, redesigns, or model downgrades. More than half told us they have no formal cost-per-outcome metric. Public comparisons of models tend to start with benchmark scores and price per million tokens. The leaders we interviewed described weighing more than that, from security and reliability to what completed tasks actually cost. At @G2dotcom, we think about AI like electricity — it’s something we want every employee to have, but we maintain a meter that tells us what usage looks like. We stopped rewarding raw token volume (tokenmaxxing) and started measuring token efficiency instead. Our AI Efficiency Index shows our internal teams who are getting the most out of every token, and we also track how fast we deliver outcomes against our objectives alongside that spend. We’re still learning, but our adoption of AI has grown to ~90% of employees using AI daily, and our token efficiency has climbed even as total spend increased. My view is that the teams that scale well will keep adoption and efficiency moving in parallel, and that and that a cost-per-outcome number is what will help them get there. Do you feel like you have to compromise on AI adoption to rein in token costs? Read the full report on how enterprise leaders really buy LLMs: research-hub.g2.com/enterpri…

Happy Homecoming, Buff Nation! 🦬 @CUBuffsFootball vs. Texas Tech at Folsom Field Presented by @ElevationsCU #GoBuffs | #ForeverFolsom

135 manufacturer websites reviewed for the Easier to Sell Index. 83% have real product data: specs, filtering, images. 99% offer no guided selling, no natural language search, no quoting for B2B.

we are so back #cpq

Every leader I talk to is trying to reshape their business to be AI-first. They know that buyers now ask AI which software to shortlist, and that agents are becoming the primary source of web traffic. They also know AI search visibility is critical to their success. How those things affect their brand, their budget, and their teams is still up for debate. Nobody has all the answers, but on November 10, @G2dotcom is bringing together a room of CMOs and senior marketing leaders in San Francisco for Reach 2026 to discuss those questions. As AI reshapes how software gets discovered, evaluated, and chosen, trust is what helps you stand out — whether a human is making the call, or an agent is doing it for them. Here's what I'm most looking forward to sharing on stage: 🎯 𝗧𝗵𝗲 𝗻𝗲𝘄 𝘁𝗿𝘂𝘀𝘁 𝗹𝗮𝘆𝗲𝗿 𝗳𝗼𝗿 𝗕𝟮𝗕 𝘀𝗼𝗳𝘁𝘄𝗮𝗿𝗲. 𝗢𝗻𝗲 𝗮𝗱𝗺𝗶𝗻, 𝗳𝗼𝘂𝗿 𝗽𝗹𝗮𝘁𝗳𝗼𝗿𝗺𝘀: Expanded advertising and buyer intent tools, plus a unified experience across G2, Capterra, Software Advice, and GetApp. 🤖 𝗜𝗻𝗻𝗼𝘃𝗮𝘁𝗶𝗼𝗻𝘀 𝗳𝗼𝗿 𝘁𝗵𝗲 𝗮𝗴𝗲𝗻𝘁𝗶𝗰 𝗲𝗿𝗮: Three new G2 products built for how buying actually happens now — Agent Evaluations, AI Blueprints, and Content Studio. 🤝 𝗩𝗼𝗶𝗰𝗲𝘀 𝗼𝗻 𝘁𝗵𝗲 𝘀𝘁𝗮𝗴𝗲: Alongside our own leadership, we’re bringing in industry leaders from @OpenAI, @semrush, and @clay to share how they're navigating this shift themselves. Software buying keeps changing. Trust doesn't. That's what G2 has been built on for over a decade. If you're a CMO or senior marketing leader who wants in the room, request your invite at reach.g2.com.

Just back from an exhilarating week in Bangalore and Gurgaon, visiting our growing @G2dotcom team and customers! Four things I'm bringing home: 𝟭. 𝗪𝗲'𝗿𝗲 𝗳𝘂𝗲𝗹𝗶𝗻𝗴 𝗴𝗹𝗼𝗯𝗮𝗹 𝗴𝗿𝗼𝘄𝘁𝗵 𝗳𝗼𝗿 𝗔𝗣𝗔𝗖 𝗹𝗲𝗮𝗱𝗲𝗿𝘀 𝘄𝗶𝘁𝗵 𝗔𝗜 𝗩𝗶𝘀𝗶𝗯𝗶𝗹𝗶𝘁𝘆. We learned from amazing customers, including Sourabh Deorah, CEO of AdvantageClub.ai, who shared how his team now ships 5x the content with AI while leveraging 876 (4.8/5) authentic G2 customer reviews to help buyers choose with confidence. GoComet grew 40% in two years, with 30% of revenue now from the US, and Johny Stephen credited G2 for showcasing them as a TMS Leader against global giants. Arwinder Kaur's Grazitti team uses G2 Buyer Intent to find buying committees and accelerate deals. As Saumitra Pandey of i-exceed put it, "AI has to discover you, it has to reason, it has to trust you." 𝟮. 𝗪𝗲 𝗮𝗿𝗲 𝗻𝗼𝘄 𝗼𝗻𝗲 𝗴𝗹𝗼𝗯𝗮𝗹 𝘁𝗲𝗮𝗺. Seven months after G2 and Gartner Digital Markets (Capterra, GetApp, Software Advice) came together, it felt like truly one team. Priti Patel reminded us how we got here: with one weekend to move everyone to G2 employment in early February, we had to "come together and be as human as we could" to embrace each other in very uncertain times. We opened our beautiful new G2 office in Gurgaon, where we can feel our vibrant PEAK culture. 𝟯. 𝗔𝗜 𝗶𝗻𝗻𝗼𝘃𝗮𝘁𝗶𝗼𝗻 𝗼𝗳𝗳 𝘁𝗵𝗲 𝗰𝗵𝗮𝗿𝘁𝘀 𝗶𝗻 𝗜𝗻𝗱𝗶𝗮. 86-90% of employees in India trust AI to boost productivity, versus 40-45% in the US. You feel it in every room and at G2. Our engineers, led by Deepan Kumar and Naman Kashyap cut PR cycle time from 190 hours to 25, leveraging our coding agents. Our SMB reps, led by @Saby Khanduja, took meeting prep from 3 hours to under 45 minutes and resolved 75% of customer cases with AI. 𝟰. 𝗣𝗘𝗔𝗞 𝗲𝗻𝘁𝗿𝗲𝗽𝗿𝗲𝗻𝗲𝘂𝗿𝗶𝗮𝗹 𝘀𝗽𝗶𝗿𝗶𝘁 𝗶𝘀 𝘁𝗵𝗿𝗶𝘃𝗶𝗻𝗴 𝗮𝘁 𝗚𝟮 𝗔𝗣𝗔𝗖. Our Market Research and Data Ops team, mostly non-engineers, built four AI agents in four days to accelerate our G2 marketplace building and accuracy. Thank you, Shreyans Dugar and Sharath Jagadish — and the entire G2 APAC team for creating an incredible week, and excited to climb to our next PEAK together!