Anthropic says it is forming an AI services company with Blackstone, Hellman & Friedman, and Goldman Sachs to work directly with mid-sized companies across industries. The key signal is not just the size of the partnership. It is that Anthropic is openly treating enterprise AI deployment as a services and implementation problem, not just a model-access problem.

If AI is going to land in real work, someone still has to identify the right workflows, implement the systems, and stay close enough to keep them running inside actual operations. That is a meaningful business-model shift. Frontier labs no longer want to be capability vendors only. They also want the delivery margin around adoption.

Perplexity is not just improving the model behind Computer. It is pushing the product into Microsoft Teams, Snowflake, Databricks, shared Workflows, and reusable Space skills. That is a clear product move from experimental assistant toward embedded team system.

What matters is where the agent is now sitting. Instead of living outside the workflow, it is moving into the collaboration layer, the data layer, and the repeatable-job layer. Once workflow templates and shared skills appear, AI execution starts to look less like a personal chat box and more like infrastructure for a team.

PwC says it is building an “AI-native finance function” with OpenAI, starting with a procurement agent inside OpenAI’s own finance organization and expanding into planning, forecasting, reporting, treasury, tax, and close workflows. The signal here is that agents are being commercialized by function, not as generic assistants for everyone.

That matters because the next phase of enterprise AI may not be one agent for the whole company. It may be domain bundles with their own controls, templates, knowledge, and review loops. Finance makes sense as an early wedge because the workflows are clear, the ROI is measurable, and the governance burden is high enough to justify packaged deployment.

IBM is positioning watsonx Orchestrate as a control plane for the multi-agent era: IBM is framing the next generation of watsonx Orchestrate as an agentic control plane with policy enforcement, accountability, sovereignty controls, and real-time data context. Once agents start multiplying inside organizations, orchestration and governance become a category of their own.

IBM Consulting and AWS are packaging a deployment platform for governed agents: Enterprise Advantage on AWS adds Context Studio, an MCP gateway, orchestration, and observability to put agents into production through a more repeatable operating model. Consulting-led deployment is being productized into a platform, not just sold as bespoke project work.

ServiceNow and Google Cloud want every agent to pass through a governed registry: AI Control Tower, integrated with Gemini Enterprise Agent Platform, gives IT and security teams visibility into which agents are running, what they touch, and how they behave. In enterprise AI, an inventory layer for agents may become as important as asset inventory in cloud or security stacks.

Microsoft Agent 365 shows the control-plane story now has real usage traction: Microsoft says tens of millions of agents have already appeared in the Agent 365 Registry after preview, with more than 500,000 agents inside Microsoft itself. It is no longer new enough to be the main spotlight, but it is still a strong signal that the control-plane layer is moving beyond slideware.

If you want agents to do real work, one of the most frustrating bottlenecks is not prompting. It is tool access. An agent can write beautifully, but if it cannot reliably touch Gmail, Drive, Docs, Sheets, or Calendar, it still sits outside the workflow. gog solves that by turning Google Workspace into a unified terminal surface for both humans and agents.

The practical value is that it removes a lot of glue work. Instead of wiring each API one by one, managing separate auth flows, and then writing wrappers so agents can use them, you get one consistent command surface for reading email, finding Drive files, opening Docs, checking Sheets, or touching Calendar.

The best way to use it is to start with one small but repetitive workflow.

For example: find a new client email, open the related Drive file, summarize the relevant Docs content, and push the output into a working checklist. Once that flow is stable, then expand into more complex tasks. That keeps auth manageable while giving the agent a clear tool layer from day one.

The impact is that you stop redesigning the stack every time you want AI to touch Google Workspace. Instead of spending time on API glue, you can spend it on the harder part: objectives, rules, approval gates, and verification.

That’s 5 sales in 4 months. 28 all time. And the performance?

14.6%, 17.6%, and 17.8%, net annualized returns on sold works held longer than one year (See all 28 at Masterworks.com).

It’s not from stocks, private equity, or real estate… it’s from contemporary and post war art. Crazy, right?

With Masterworks, you don’t need to be a BILLIONAIRE to invest in multi-million dollar art anymore.

Historically, the segment overall has had attractive appreciation and low correlation to stocks.*

Masterworks targets works featuring legends like Banksy, Basquiat, and Picasso, identifying what they believe to have significant long-term appreciation potential, not just at the artist level but at the level of individual artworks.

As one of the largest players in the art market, with $1.3 billion invested over 500 artworks, they pass critical advantages through to their 70,000+ members to add art to their portfolios strategically.

Looking to diversify your investments in 2026?

Click here to skip the waitlist

^{*According to Masterworks data. Investing involves risk. Past performance is not indicative of future returns. See important Reg A disclosures at masterworks.com/cd.}

Wharton’s blueprint breaks agent-adoption friction into three practical layers: perceived competence, trust, and delegation of control. Companies are not just asking what an agent can do. They are asking whether they are willing to hand it work, whether they can see what it is doing, and whether they can stop it at the right moment.

Implication: That is exactly why service layers, governance layers, and human-in-the-loop design are expanding so quickly around enterprise agent deployment.

Anthropic’s report says more than half of businesses are already deploying multi-step agent workflows, and a meaningful share are moving into cross-functional processes. Once agents stop handling isolated tasks and start touching longer chains of work, the need for observability, handoffs, verification, and runtime support rises almost by default.

Implication: If adoption maturity is shifting from single-step assistants to workflow agents, the next budget winner may be orchestration and rollout support, not just the model layer.