|
The Agentic Enterprise
AK · 7 min read
|
|
Thursday, July 16, 2026
The model makers want to run your rollout.
Anthropic and Blackstone launched Ode, a $1.5 billion firm that embeds engineers inside companies to put AI to work. OpenAI, Deloitte, and Accenture are building the same thing.
The frontier labs have quietly conceded the point every CIO already knew: the hard part of enterprise AI was never the model, it was making it work inside a real company. Now the people who sell you the model want to sell you the team that installs it, and a new $1.5 billion company is betting the last mile is where the trillion-dollar business actually is. The catch for buyers is who those installers answer to.
|
|
Ode, and the race to own the last mile of enterprise AI.
| T |
wo of the biggest names in AI just agreed the model is the easy part. Ode with Anthropic launched this week as a $1.5 billion company built to embed engineers inside enterprises and rewire their core processes around AI. Anthropic, Blackstone, Hellman & Friedman, and Goldman Sachs supplied the capital; the firm runs on Fractional AI, the boutique it absorbed, and roughly 100 engineers. |
The launch is a concession. For two years the labs sold capability and assumed adoption would follow. It has not, at least not at the pace the valuations imply. Ode's own pitch is that non-AI companies win only if they adopt the technology the right way, which takes talent almost none of them have. OpenAI reached the same conclusion with its Deployment Company. Deloitte and Accenture built forward-deployed engineering practices before either lab did.
For a CIO, this reshapes the build-buy-services question. The vendor selling you the model now wants the integration contract too, and it arrives with a "Claude-first" default and private-equity owners who will steer their portfolio companies to it as customers. That is convenient and conflicted in the same breath. The scarce input is people: everyone is chasing the same small pool of elite applied-AI engineers, so "special forces, not an army" is a positioning line and a supply constraint at once. Ode says it will scale internationally while staying boutique, the needle every services firm claims it can thread and few do.
|
The hard part was never the model. It was your company.
|
|
The Spearhead Take
Treat a lab-owned integrator like any other vendor with a house preference: fast, capable, and not neutral. Keep model selection, your evals, and the orchestration layer in your own hands, scope the work to a business outcome you can measure, and make sure the know-how those engineers build stays in the building after they leave.
|
|
|
The Obvious & The Overlooked
Three reads the market has. Four it is missing.
|
The Obvious
The labs concede that models alone will not drive adoption.
Anthropic and OpenAI each stood up a separate services arm to do the installing. TechCrunch
The consultancies saw it first.
Deloitte and Accenture built forward-deployed engineering practices before either lab did. TechCrunch
Private equity wants its portfolio AI-ified.
Blackstone conceived Ode after its own portfolio companies kept stalling on adoption. TechCrunch
|
The Overlooked
"Claude-first" is a conflict, not a feature.
An integrator part-owned by a model vendor has a thumb on the model-selection scale. TechCrunch
Your integrator's owners may also own your rivals.
The private-equity backers will funnel their portfolio companies to Ode as customers. TechCrunch
The bottleneck is people, not models.
Demand for elite forward-deployed engineers already outstrips supply, by the venture's own account. TechCrunch
Nadella just named the hidden cost.
The deeper an integrator goes, the more proprietary know-how leaves your building. The Register
|
|
|
Moving Pieces
Five developments worth a CIO's attention.
Deployment
Cisco hands an AI agent to all 90,000 employees
Cisco will give every one of its roughly 90,000 employees a personal AI agent when its new fiscal year opens at the end of July, with each agent routing tasks to whichever model is cheapest rather than defaulting to a frontier one. CFO Mark Patterson says AI already drafts 80 to 90 percent of the first pass of the company's MD&A filings. The enterprise read: this is one of the largest all-hands agent rollouts at a named Fortune 500, and the cost-routing design is the tell. The frontier model is becoming a fallback, not the default, and it arrives the same month Cisco is trimming jobs.
Policy
The FTC opens a comment window on AI accuracy
The Federal Trade Commission is taking public comment through July 31 on a policy statement addressing state laws that would require companies to alter AI model outputs. It is a small filing with a large signal: the federal posture on AI is hardening from voluntary ethics language into enforceable duties, and the fight over who can compel changes to a model's answers is now live. The enterprise read: your first real compliance test is less likely to come from a regulator than from your own procurement and legal teams asking who controls the output, how it is logged, and who is liable when it is wrong.
Infrastructure
SnapLogic ships governed plumbing for AI agents
SnapLogic made SnapCode and its MCP Server generally available, connecting AI agents to enterprise systems through more than 1,000 prebuilt connectors for SAP, Oracle, Salesforce, Snowflake, Workday, ServiceNow, and mainframe environments, with central security and monitoring on every integration. The enterprise read: the interesting layer this year is not the agent, it is the governed pipe between the agent and your systems of record. Whoever owns that integration and control plane owns the audit trail regulators and boards will eventually ask to see, which makes this plumbing a strategic choice, not a procurement afterthought.
Deals
Nous Research nears $75M at a $1.5B valuation for an open-source agent
Nous Research is finalizing a round of at least $75 million led by Robot Ventures at a $1.5 billion valuation, a sharp step up from the $1 billion mark it hit a year ago. The draw is Hermes, its open-source agent, which carries roughly 214,000 GitHub stars alongside a cloud-hosted tier. The enterprise read: the open-weight camp now has a well-funded flag-bearer aimed squarely at teams that want to run agents on their own infrastructure rather than rent them. For any leader worried about vendor lock-in or data leaving the building, an open-weight option with real backing is worth a line in the evaluation.
Product
Kore.ai launches Artemis to take on Salesforce and ServiceNow
Kore.ai released Artemis, an agent platform pitched at enterprises building and governing their own agents, positioning the mid-tier vendor directly against Microsoft, Salesforce, and ServiceNow. The enterprise read: the agent-platform field is crowding fast, and the incumbents no longer have it to themselves. That is good for pricing and leverage, but it also means the platform you standardize on this year may look very different, or be owned by someone else, in eighteen months. Favor platforms that keep your agents and data portable over the ones that promise the tightest single-vendor bundle.
|
|
On the Radar
Eight signals, sharpened.
| Product |
ServiceNow extended its Autonomous Workforce across every major business function. The governance-and-control pitch is now the incumbents' main defense against nimbler agent startups. ServiceNow |
| Product |
OpenAI folded its desktop apps into Chat, Codex, and Work, making agents the default surface. The consumer-to-enterprise funnel is being redrawn around agentic workflows rather than a chat box. OpenAI |
| Security |
OpenAI expanded its Daybreak program with GPT-5.5-Cyber and a "Patch the Planet" push to fix open-source bugs. AI security is shifting from finding flaws to shipping verified fixes at scale. OpenAI |
| Policy |
The EU published a July 2026 action plan on cybersecurity and AI to help members and businesses defend advanced models. Governance is widening from what models may do to how they must be secured. European Commission |
| Deployment |
Gartner projects 40% of enterprise applications will ship with embedded agents by year-end, up from under 5% in 2025. The default software experience is becoming agentic whether or not you have a strategy for it. AI News |
| Governance |
Microsoft reports 80%-plus of the Fortune 500 now run active AI agents, but only about 11% at true production scale. The gap between piloting and shipping remains the real enterprise story. Microsoft |
| Deals |
AIsphere closed a $439 million Series C led by Alibaba for generative AI video. Video generation is drawing megacap-scale checks even as enterprise use cases stay early. Scouts by Yutori |
| Deals |
Chai Discovery raised a $400 million Series C for AI drug discovery, backed by Index, Kleiner Perkins, and Sequoia. Applied AI in regulated R&D is now a top-tier venture category, not a science project. Scouts by Yutori |
|
|
Quick Hits
The wider field, one line each.
| SAP is acquiring Freiburg's Prior Labs and will invest more than $1.18 billion to build a European frontier AI lab around tabular foundation models. TechCrunch |
| Thira raised $21 million in seed funding led by Madrona, founded by Apptio's co-founders, to build enterprise AI. Scouts by Yutori |
| TYLsemi emerged from stealth with a $43 million round led by Matter Venture Partners for custom AI chip design. Scouts by Yutori |
| Harvey reportedly raised a $200 million Series C at a $2.1 billion valuation for legal AI. AI Funding |
| Glean reportedly raised a $180 million Series D at a $2.7 billion valuation as enterprise search consolidates. AI Funding |
| Hebbia reportedly raised a $130 million Series B at a $1 billion valuation for AI document analysis in finance. AI Funding |
| Lovable reportedly raised a $200 million Series B at a $2.8 billion valuation for AI app building. AI Funding |
| AI-agent startups raised more than $1.8 billion across a dozen-plus deals in July, led by enterprise automation and developer tools. AI Funding |
| Fractional AI, the boutique now powering Ode, ended an 11-month OpenAI partnership when the joint venture acquired it. TechCrunch |
| Meta pulled a controversial Instagram AI feature after user backlash. TechCrunch |
|
|
The Number
86¢
AI's share of US venture dollars, H1 2026
Of every dollar US venture capital invested in the first half of 2026, that much went to AI.
PitchBook put US venture deal value at $412.7 billion for the six months, nearly 30 percent above all of last year, and AI companies took $355.9 billion of it. Capital this concentrated builds fast and breaks hard. The vendors your roadmap depends on are being funded by the narrowest wave in venture history, which is a dependency worth naming before you standardize on any single one of them.
|
|
Counter-Signal
Governance
The deeper AI goes, the more of your company it takes with it.
The same week the labs pitched embedding their engineers inside your operations, Microsoft's own CEO warned about the cost. Satya Nadella called it the reverse information paradox: you pay for intelligence twice, once in tokens and again in the proprietary knowledge you must reveal to make the model useful. Models learn from the exhaust, he wrote, the prompts, the tool calls, and above all the corrections your people make when the system is wrong. Every correction is institutional know-how, distilled and carried off.
That is the quiet tension under the deployment gold rush. An engineer sent to rewire your most important business process is, by design, learning it in detail, and so is the model behind them. Nadella's prescription is worth stealing whoever you hire: build private evaluations, keep the orchestration layer independent of any single vendor, and hold your data and learning loop inside your own network. Convenience flows to the vendor by default. Keeping the knowledge is now a deliberate act of engineering.
|
|
From the Field
Every technology wave eventually rediscovers the consultant.
We spent a decade being told software would eat services. Buy the platform, self-serve, scale without bodies. Then the most valuable software of this era turned out to need more hands to install than anything before it, and the model makers, of all people, are the ones staffing up to do the installing. Ode calls its engineers special forces. OpenAI has a Deployment Company. Deloitte and Accenture have forward-deployed practices. The org chart of the AI boom is starting to look a lot like 2005, only with better tools.
That is not a criticism. The work is real. Turning a hallucinating, general-purpose model into a system that survives your procurement process, your compliance review, and your ordinary Tuesday is genuinely hard, and most companies do not have the people to do it. Buying that talent by the sprint is a reasonable move. The trap is forgetting that the people who install the system also learn your business while they do it, and that the ones sent by your model vendor arrive with a view on which model you should use.
So take the help. Just keep the parts that make you you, the evals, the data, the judgment about what good looks like, on your side of the table. The last mile is where the value is. It is also where the dependency is. Make sure you own the map, even when someone else is driving.
Let's get to production,
AK
|
|
|
|
Anthropic is a Spearhead technology partner, and its Claude model produced this edition under human editorial direction. The partnership earned no favorable framing: the Big Story on Ode with Anthropic is handled with harder scrutiny, flagging the "Claude-first" conflict of interest, the private-equity owners steering captive customers, and the elite-engineer supply constraint, and it applies the same critical frame to OpenAI's Deployment Company and to Deloitte's and Accenture's forward-deployed practices. The Ode figures ($1.5B valuation, roughly 100 engineers, the Fractional AI acquisition) are from TechCrunch's July 15 exclusive. Cisco's ~90,000-employee agent rollout and 80-90% MD&A drafting figure are from Fortune; the FTC comment window (through July 31) is from the FTC; SnapLogic's SnapCode and MCP Server GA are from GlobeNewswire and BigDATAwire; Nous Research's ~$75M round at a $1.5B valuation is described as "finalizing," not closed, per TechCrunch and The Block; Kore.ai's Artemis launch is from VentureBeat. Satya Nadella's "reverse information paradox" is from his July 12 post via TechCrunch and The Register. The Number (US H1 2026 venture at $412.7B, $355.9B to AI, an 86% share) is from SiliconANGLE citing PitchBook and is single-source at Tier 3. Several Quick Hits funding figures (Thira, TYLsemi, Harvey, Glean, Hebbia, Lovable, the $1.8B July agent aggregate) and the AIsphere and Chai Discovery rounds on the Radar are drawn from funding-tracker aggregators and were not independently verified against primary filings; treat the valuations as reported. SAP's Prior Labs acquisition (announced in May) is carried as recent context. No emoji, hashtags, or exclamation marks were used. No India-domiciled outlets were used. All editorial decisions are human-directed.
|
|
The Agentic Enterprise
Know more about AI than 95% of your peers. By 7 AM.
A daily AI intelligence briefing for enterprise leaders, published by Spearhead. We build AI systems that work. Strategy. Engineering. Production. Outcomes.
© 2026 Spearhead. All rights reserved.
|
|
