How Enterprises Startups Benchmark AI Maturity
James Lamberti · August 13, 2025
venture capitalstart upsbranding
Georgian's AI Applied Benchmarks report segments companies into Crawlers, Walkers, Joggers, and Runners, with runners defined by breadth of use cases, depth (scaling pilots into production), and tying AI to clear ROI. The study finds enterprises are adopting AI as fast or faster than growth-stage firms in many areas (support, data analytics, personalization), with top concerns being quality, integration, security, accountability, and upskilling. Overall sentiment is positive and the report invites community participation in future waves.
0:05 — David: Hey everyone and welcome to another edition of AI Marketers Guild. I joked earlier about K-pop Demon Hunters, but we’ll have a lot of fun. I’m joined by an old friend, James Lamberti, whom I’ve known for decades. Many of you took part in research fielded through the VC firm Georgian, and James has results to share.
1:10 — James: Thanks — great to be here. I’ve been a CMO, head of growth, and GM across about 11 growth-stage companies. A little over a year ago I joined Georgian, a growth equity VC firm based in Canada with roughly 50 portfolio companies and about $5–6B AUM. My mandate is to work with 50+ CEOs on go-to-market challenges and run targeted secondments.
2:15 — James: Georgian differentiates itself with a 20-person, PhD- and VP-level AI lab launched in 2018. That lab is a major part of what makes the firm unique and is the backdrop for this work.
3:19 — James: The AI Applied Benchmarks report is an open research project. We invited a community of thought leaders — including AI Marketers Guild, go-to-market partners, the Vector Institute, academics, and other VCs — to contribute. The program runs in Tel Aviv, San Francisco, New York, Toronto, Montreal, and London.
4:22 — James: You can find the reports at georgian.io. Today we’re discussing the enterprise vs. growth-stage view. Methodologically, the study is rigorous — run through a third party (NewtonX), quantified, and intended to cut through hype and show what’s actually happening.
5:23 — James: We benchmark AI application across companies and will walk through the data today. We hope to involve this community more in future waves.
6:26 — David: We started with a formal segmentation: Crawlers, Walkers, Joggers, and Runners. For this audience, self-evaluate where you are. Quick question to kick off: what factors define a runner?
7:34 — David: My hot take — runners learn by doing: they pilot and implement, learn from failures, and emphasize cross-functionality. It’s not just a single team improving; it ladders up to broader organizational goals. Runners have methods for disseminating learning across teams so staff must learn these skills to stay relevant.
8:39 — David: We ran a poll — the room skewed toward joggers (high intermediate maturity), then walkers. Even in a sophisticated crowd, many self-evaluate in the middle of the curve. People here might be hard graders, but the distribution is typical.
9:42 — James: I’d add three defining factors for runners. One: breadth — AI deployed across many functions from legal, operations, and customer success to marketing, sales, and R&D. Two: depth — moving from individual “acts of heroism” to scaling pilots into production, with a train-the-trainer mentality. Three: focus on ROI — tying initiatives to enterprise KPIs (cost savings or ARR) strongly correlates with being a runner.
10:47 — David: A useful point from the chat — there’s often a discrepancy between individual usage and organizational adoption. Individuals may use AI, but if it’s not institutionalized it won't move a company from walker to runner.
11:48 — David: How do growth-stage and enterprise companies map to these segments?
12:49 — David: There are myths about nimble startups (kayaks) vs. big enterprises (ocean liners). With this wave of AI, many of those assumptions are false. Enterprises have adopted certain AI capabilities very quickly; the differences are smaller or even the opposite of expectations.
13:51 — James: I agree. I’ve never seen enterprises adopt technology this quickly. Historically growth-stage companies led adoption, but not with this AI wave. Many large companies adopted rapidly and sometimes even before they fully understood it; that’s useful experimentation.
14:53 — James: Next poll — where do you have AI in production? By production, I mean scaled, enterprise-wide deployment, not just a single person’s tool.
15:57 — David: Poll results show enterprises and growth-stage firms are eerily similar across many functions. Support/chatbots, data analytics, and personalization show strong adoption; cybersecurity underindexes slightly for this sample.
17:01 — David: In a marketing-heavy room, support and chatbots overindex. Chatbots have been experimented with for years, so it’s natural to see more adoption there. Data readiness and analytics remain hard but present bigger opportunities in enterprises.
18:02 — James: The enterprise’s need for AI is more acute because of scale and fragmented data. AI can let teams “converse” with data even when it’s unstructured and scattered.
19:02 — James: New poll — which go-to-market use cases are you piloting or using? Think organized pilots or scaled rollouts across GTM functions.
20:03 — Participant: From my conversations, sales and BDR use cases top the list because of immediate impact: lead scoring, segmentation, and integrating AI with CDPs, analytics, and marketing tools to get insights from connected data.
21:09 — Participant: Customer support is also strong. For example, AI can analyze a client’s website and recommend code changes or support items in minutes; that used to take hours or days from a human.
22:10 — David: Are you succeeding in moving from individual acts of productivity to full-scale production and train-the-trainer models?
23:10 — Participant: Yes. For developer-heavy companies the developers can be skeptical, but the rest of the team often embraces anything that increases productivity and reduces load. Enterprises are starting with developer-centric use cases, especially in R&D, but scaling from developer productivity to product-level success is tougher than go-to-market scaling.
24:14 — James: The benchmarks separate two “swim lanes”: an office-of-the-CTO technical view and a go-to-market view. Converting developer productivity gains into scaled product improvements is a big challenge.
25:18 — James: Back to GTM specifics — marketing messaging and content is the highest use case across both enterprise and growth-stage. Why does enterprise sometimes overindex on content versus startups?
26:22 — David: One reason is enterprises have concerns about training models on proprietary content and over-relying on AI-generated content. But many companies are pragmatic and willing to use AI for rapid output when appropriate.
27:28 — James: I disagree that AI will necessarily degrade content quality. Enterprises have rich private data and original content; they can combine proprietary knowledge with public data to produce higher-quality content. Startups, by using brand guidelines, persona work, and objection handling codified into agents, can achieve consistent, high-quality messaging very quickly.
28:29 — Participant: Yes — startups can get sloppy with branding, but with AI-driven brand codification they can reach a solid level quickly. Enterprises may prefer human creativity and will sometimes bend brand rules, which can be better or worse depending on the case.
29:29 — James: For many startups, formalizing messaging with AI frees them to focus on their core business while keeping brand quality consistent.
30:28 — David: Let’s move to RevOps and lead scoring/segmentation. How will RevOps change in one to two years because of AI?
31:28 — Participant: RevOps needs clear data governance. If teams share data properly, RevOps can provide the assets (messaging, one-pagers, vertical content) to sales without constant back-and-forth. That speeds sales and reduces friction.
32:25 — David: AI will change RevOps by making data more conversational and reducing the need for a large stack of tooling. We’re already seeing divestment from some legacy marketing tech in favor of a thin, high-quality CRM/martech layer plus conversational AI on top.
33:30 — James: Good RevOps depends on a clean data foundation and the ability to trust inputs. A well-trained agent on reliable data can become the foundation for RevOps, transforming fragmented data into usable insights.
34:32 — Participant: From an analytics background, we’ve over-engineered many solutions. AI forces us to downscale to what matters and organize data for cross-team use. Human insight still matters to define what to capture; AI helps enforce and surface it.
35:39 — James: Next poll — how do you measure AI product initiatives? Options: direct impact on new revenue, direct impact on cost savings, tied to cost of revenue, indirect benefit, or unsure.
36:36 — David: Poll results in this room favored direct impact on revenue and direct impact on cost savings. Tying AI initiatives explicitly to revenue or cost savings correlates with scaling and showing value.
37:35 — James: That does mark a runner behavior: tying initiatives to clear KPIs. But some AI investments are more like fire insurance or strategic capability-building, where ROI is longer-term or indirect. The cost of falling behind is also a factor.
38:30 — David: Many organizations are planting seeds now by investing to learn, even if immediate ROI is unclear. That learning can pay off quickly in the next cycles.
39:27 — James: Next poll — top concerns about AI adoption (pick three). Common answers: quality of insights, integration with existing systems, data security, accountability, upskilling teams, brand standards, and maintenance.
40:33 — David: The poll results mirrored the study: quality, integration, and security are top concerns, along with upskilling and accountability.
41:40 — James: The concern about quality puzzles me a bit. In enterprises, with proprietary data and good governance, AI can maintain or improve quality. Many quality issues today stem from early experimentation, bad inputs, or failing to include a human-in-the-loop.
42:42 — Participant: Hallucinations are real but solvable. They’re similar to human errors or bad SQL queries. Policies, quality control, and experienced analysts can catch and fix these issues quickly.
43:50 — James: The human-in-the-loop is critical. Programmatic content shouldn't be your entire strategy without governance. When systems are structured with RAG, sanitized data, and guardrails, hallucinations decline.
44:53 — James: Final poll — how has AI impacted your organization: positive, neutral, or negative? Results: generally very positive impact, slightly higher in growth-stage companies which are resource-constrained and move quickly.
45:58 — James: We’re sharing a lot of data in the full benchmark report — I encourage you to download it. Waves 1 and 2 are available at georgian.io; waves 3 and 4 are coming in Q1 and Q3 of 2026. We want this community involved in future waves.
47:03 — David: We’ll host a session with James and Sram Vajre in October to look at what go-to-market teams might look like in 2028 and to explore emerging technologies.
48:02 — James: Waves three and four will include more community participation. We’ll make it easy for AI Marketers Guild members to contribute and to get a community-specific cut of the data.
49:00 — David: Thanks, James — great session. We’ll share the report and follow up with links and next steps. Appreciate your time.
50:00 — James: Thanks everyone. Have a great Labor Day weekend. Look forward to more sessions and future participation.
51:01 — David: Thanks again, James. Appreciate it.
(End of transcript)
0:05 — David: Hey everyone and welcome to another edition of AI Marketers Guild. I joked earlier about K-pop Demon Hunters, but we’ll have a lot of fun. I’m joined by an old friend, James Lamberti, whom I’ve known for decades. Many of you took part in research fielded through the VC firm Georgian, and James has results to share.
1:10 — James: Thanks — great to be here. I’ve been a CMO, head of growth, and GM across about 11 growth-stage companies. A little over a year ago I joined Georgian, a growth equity VC firm based in Canada with roughly 50 portfolio companies and about $5–6B AUM. My mandate is to work with 50+ CEOs on go-to-market challenges and run targeted secondments.
2:15 — James: Georgian differentiates itself with a 20-person, PhD- and VP-level AI lab launched in 2018. That lab is a major part of what makes the firm unique and is the backdrop for this work.
3:19 — James: The AI Applied Benchmarks report is an open research project. We invited a community of thought leaders — including AI Marketers Guild, go-to-market partners, the Vector Institute, academics, and other VCs — to contribute. The program runs in Tel Aviv, San Francisco, New York, Toronto, Montreal, and London.
4:22 — James: You can find the reports at georgian.io. Today we’re discussing the enterprise vs. growth-stage view. Methodologically, the study is rigorous — run through a third party (NewtonX), quantified, and intended to cut through hype and show what’s actually happening.
5:23 — James: We benchmark AI application across companies and will walk through the data today. We hope to involve this community more in future waves.
6:26 — David: We started with a formal segmentation: Crawlers, Walkers, Joggers, and Runners. For this audience, self-evaluate where you are. Quick question to kick off: what factors define a runner?
7:34 — David: My hot take — runners learn by doing: they pilot and implement, learn from failures, and emphasize cross-functionality. It’s not just a single team improving; it ladders up to broader organizational goals. Runners have methods for disseminating learning across teams so staff must learn these skills to stay relevant.
8:39 — David: We ran a poll — the room skewed toward joggers (high intermediate maturity), then walkers. Even in a sophisticated crowd, many self-evaluate in the middle of the curve. People here might be hard graders, but the distribution is typical.
9:42 — James: I’d add three defining factors for runners. One: breadth — AI deployed across many functions from legal, operations, and customer success to marketing, sales, and R&D. Two: depth — moving from individual “acts of heroism” to scaling pilots into production, with a train-the-trainer mentality. Three: focus on ROI — tying initiatives to enterprise KPIs (cost savings or ARR) strongly correlates with being a runner.
10:47 — David: A useful point from the chat — there’s often a discrepancy between individual usage and organizational adoption. Individuals may use AI, but if it’s not institutionalized it won't move a company from walker to runner.
11:48 — David: How do growth-stage and enterprise companies map to these segments?
12:49 — David: There are myths about nimble startups (kayaks) vs. big enterprises (ocean liners). With this wave of AI, many of those assumptions are false. Enterprises have adopted certain AI capabilities very quickly; the differences are smaller or even the opposite of expectations.
13:51 — James: I agree. I’ve never seen enterprises adopt technology this quickly. Historically growth-stage companies led adoption, but not with this AI wave. Many large companies adopted rapidly and sometimes even before they fully understood it; that’s useful experimentation.
14:53 — James: Next poll — where do you have AI in production? By production, I mean scaled, enterprise-wide deployment, not just a single person’s tool.
15:57 — David: Poll results show enterprises and growth-stage firms are eerily similar across many functions. Support/chatbots, data analytics, and personalization show strong adoption; cybersecurity underindexes slightly for this sample.
17:01 — David: In a marketing-heavy room, support and chatbots overindex. Chatbots have been experimented with for years, so it’s natural to see more adoption there. Data readiness and analytics remain hard but present bigger opportunities in enterprises.
18:02 — James: The enterprise’s need for AI is more acute because of scale and fragmented data. AI can let teams “converse” with data even when it’s unstructured and scattered.
19:02 — James: New poll — which go-to-market use cases are you piloting or using? Think organized pilots or scaled rollouts across GTM functions.
20:03 — Participant: From my conversations, sales and BDR use cases top the list because of immediate impact: lead scoring, segmentation, and integrating AI with CDPs, analytics, and marketing tools to get insights from connected data.
21:09 — Participant: Customer support is also strong. For example, AI can analyze a client’s website and recommend code changes or support items in minutes; that used to take hours or days from a human.
22:10 — David: Are you succeeding in moving from individual acts of productivity to full-scale production and train-the-trainer models?
23:10 — Participant: Yes. For developer-heavy companies the developers can be skeptical, but the rest of the team often embraces anything that increases productivity and reduces load. Enterprises are starting with developer-centric use cases, especially in R&D, but scaling from developer productivity to product-level success is tougher than go-to-market scaling.
24:14 — James: The benchmarks separate two “swim lanes”: an office-of-the-CTO technical view and a go-to-market view. Converting developer productivity gains into scaled product improvements is a big challenge.
25:18 — James: Back to GTM specifics — marketing messaging and content is the highest use case across both enterprise and growth-stage. Why does enterprise sometimes overindex on content versus startups?
26:22 — David: One reason is enterprises have concerns about training models on proprietary content and over-relying on AI-generated content. But many companies are pragmatic and willing to use AI for rapid output when appropriate.
27:28 — James: I disagree that AI will necessarily degrade content quality. Enterprises have rich private data and original content; they can combine proprietary knowledge with public data to produce higher-quality content. Startups, by using brand guidelines, persona work, and objection handling codified into agents, can achieve consistent, high-quality messaging very quickly.
28:29 — Participant: Yes — startups can get sloppy with branding, but with AI-driven brand codification they can reach a solid level quickly. Enterprises may prefer human creativity and will sometimes bend brand rules, which can be better or worse depending on the case.
29:29 — James: For many startups, formalizing messaging with AI frees them to focus on their core business while keeping brand quality consistent.
30:28 — David: Let’s move to RevOps and lead scoring/segmentation. How will RevOps change in one to two years because of AI?
31:28 — Participant: RevOps needs clear data governance. If teams share data properly, RevOps can provide the assets (messaging, one-pagers, vertical content) to sales without constant back-and-forth. That speeds sales and reduces friction.
32:25 — David: AI will change RevOps by making data more conversational and reducing the need for a large stack of tooling. We’re already seeing divestment from some legacy marketing tech in favor of a thin, high-quality CRM/martech layer plus conversational AI on top.
33:30 — James: Good RevOps depends on a clean data foundation and the ability to trust inputs. A well-trained agent on reliable data can become the foundation for RevOps, transforming fragmented data into usable insights.
34:32 — Participant: From an analytics background, we’ve over-engineered many solutions. AI forces us to downscale to what matters and organize data for cross-team use. Human insight still matters to define what to capture; AI helps enforce and surface it.
35:39 — James: Next poll — how do you measure AI product initiatives? Options: direct impact on new revenue, direct impact on cost savings, tied to cost of revenue, indirect benefit, or unsure.
36:36 — David: Poll results in this room favored direct impact on revenue and direct impact on cost savings. Tying AI initiatives explicitly to revenue or cost savings correlates with scaling and showing value.
37:35 — James: That does mark a runner behavior: tying initiatives to clear KPIs. But some AI investments are more like fire insurance or strategic capability-building, where ROI is longer-term or indirect. The cost of falling behind is also a factor.
38:30 — David: Many organizations are planting seeds now by investing to learn, even if immediate ROI is unclear. That learning can pay off quickly in the next cycles.
39:27 — James: Next poll — top concerns about AI adoption (pick three). Common answers: quality of insights, integration with existing systems, data security, accountability, upskilling teams, brand standards, and maintenance.
40:33 — David: The poll results mirrored the study: quality, integration, and security are top concerns, along with upskilling and accountability.
41:40 — James: The concern about quality puzzles me a bit. In enterprises, with proprietary data and good governance, AI can maintain or improve quality. Many quality issues today stem from early experimentation, bad inputs, or failing to include a human-in-the-loop.
42:42 — Participant: Hallucinations are real but solvable. They’re similar to human errors or bad SQL queries. Policies, quality control, and experienced analysts can catch and fix these issues quickly.
43:50 — James: The human-in-the-loop is critical. Programmatic content shouldn't be your entire strategy without governance. When systems are structured with RAG, sanitized data, and guardrails, hallucinations decline.
44:53 — James: Final poll — how has AI impacted your organization: positive, neutral, or negative? Results: generally very positive impact, slightly higher in growth-stage companies which are resource-constrained and move quickly.
45:58 — James: We’re sharing a lot of data in the full benchmark report — I encourage you to download it. Waves 1 and 2 are available at georgian.io; waves 3 and 4 are coming in Q1 and Q3 of 2026. We want this community involved in future waves.
47:03 — David: We’ll host a session with James and Sram Vajre in October to look at what go-to-market teams might look like in 2028 and to explore emerging technologies.
48:02 — James: Waves three and four will include more community participation. We’ll make it easy for AI Marketers Guild members to contribute and to get a community-specific cut of the data.
49:00 — David: Thanks, James — great session. We’ll share the report and follow up with links and next steps. Appreciate your time.
50:00 — James: Thanks everyone. Have a great Labor Day weekend. Look forward to more sessions and future participation.
51:01 — David: Thanks again, James. Appreciate it.
(End of transcript)