The Early Scale: Tuesday, July 21, 2026

Good morning

Good morning. It's Tuesday, and the story of this week is already clear: AI is everywhere, but scaling it is the hard part. Token costs are too high, governance frameworks are too thin, and enterprise buyers are now researching vendors on ChatGPT before they ever visit your website. The gap between "we deployed AI" and "AI is working for us" is where most businesses are stuck right now. Let's close that gap, starting with your next five minutes.

The Big Three

72% of B2B Buyers Use ChatGPT to Vet Vendors. Most Brands Don't Show Up.

Nearly three-quarters of B2B software buyers now use ChatGPT as part of their vendor evaluation process, according to new research cited by MarketScale. The problem: 51% of tech brands have zero AI citations in large language model outputs. If an enterprise buyer asks ChatGPT who the best vendor is in your category and your name doesn't appear, you're not in the deal.

The B2B angle: Audit your brand's LLM presence this week by querying ChatGPT, Perplexity, and Gemini with the exact phrases your buyers use to find vendors in your category, then build a GEO content strategy to close the gap.

Palo Alto CEO: Token Costs Must Fall 90% in Two Years for Enterprise AI to Scale

Nikesh Arora put a precise number on a problem every enterprise CIO already feels: AI inference costs are still too high to run at production scale. He says token prices need to drop 90% within two years for the economics to work. As evidence, Uber burned through its entire full-year AI budget by April. This isn't a niche concern for tech companies; it's a capital allocation problem for any business running AI workloads.

The B2B angle: Before expanding AI pilots to production, model out your per-token cost trajectory against projected usage volume and build a cost ceiling into your AI budget now, before you hit Uber's problem at a smaller scale.

Anthropic and Blackstone's $1.5B Venture Ode Embeds Engineers Directly Inside Enterprise Clients

Ode, a $1.5 billion joint venture backed by Blackstone, Goldman Sachs, Hellman & Friedman, and anchored by Anthropic, takes a forward-deployed engineering model into enterprise AI adoption. Instead of selling software and leaving clients to figure out implementation, Ode puts engineers physically inside client organizations to close the gap between AI deployment and AI that actually works. It's a direct acknowledgment that the implementation layer is where enterprise AI breaks down.

The B2B angle: If your AI initiatives are stalling in the pilot phase, evaluate whether you need embedded technical partnership rather than another software license, and use Ode's model as a benchmark when negotiating implementation support from any AI vendor.

Also worth knowing

Global M&A hit a record $2.8 trillion in H1 2026, up 48% year-on-year, with industrial manufacturing surging 28% as AI infrastructure, grid modernization, and defense reshape deal flow. If you're in manufacturing, industrials, or energy, your competitive landscape is being redrawn right now.

Kyndryl's 2026 People Readiness Report finds 57% of enterprises run AI in core processes, but only 23% of leaders say their workforce is fully prepared. Deployment is outrunning training, and that gap is where costly AI failures happen.

HubSpot's State of Ecosystems 2026 report projects the partner opportunity at $42 billion by 2030, driven by mid-market demand for AI operationalization help. Channel and solutions partners who build AI implementation practices now are positioned to capture a disproportionate share of that growth.

Smart plays for the week

Run your own brand through ChatGPT, Perplexity, and Gemini using the five to seven search phrases your buyers actually use, document where you appear or don't, and bring the results to your next marketing meeting as a GEO gap analysis. 72% of B2B buyers now use AI tools to evaluate vendors before ever visiting a website, and 51% of tech brands have zero citations, making LLM visibility a pipeline issue, not just an SEO issue.

Before your next AI vendor renewal or expansion decision, pull actual token usage logs for the past 90 days, project them against full-year production volume, and set a written cost ceiling in your AI budget before you hit overage. Uber burned its entire full-year AI budget by April, and Palo Alto's Nikesh Arora says current token economics still don't support production-scale deployment for most enterprises.

Audit your AI adoption roadmap against Kyndryl's 57/23 gap and schedule a workforce readiness assessment before you greenlight any new AI deployment in a core process. Kyndryl's 2026 data shows enterprises are deploying AI far faster than they're preparing workers to use it, and that gap is exactly where costly failures and poor ROI compound.

Something to think about

It's a rare moment of executive candor about the economics of AI at scale. Most enterprise AI conversations still center on capability. Arora is centering the conversation on cost, which is the actual constraint that separates pilots from production.

Teach me something: GEO: Generative Engine Optimization

GEO is the practice of structuring your content, authority signals, and data so that AI-powered search tools like ChatGPT, Perplexity, and Google's AI Overviews surface your brand in their responses. Unlike traditional SEO, which targets crawlable web pages ranked by algorithms, GEO targets the training data, citation patterns, and source preferences that large language models use when generating answers. For B2B brands, this matters because buyers are now asking AI tools to recommend vendors, compare solutions, and summarize categories before they visit a single website. If your thought leadership, product pages, and PR placements aren't structured to be cited by LLMs, you don't exist in that part of the buyer journey.