A recent McKinsey piece on "The Automation Curve" in agentic commerce has been making the rounds, highlighting a shift we see every day at VGS. We are moving away from a world of "search and click" toward a world of "intent and execution."
The McKinsey framework provides a brilliant roadmap (see below), and the reality on the ground, especially for merchants, reveals several fascinating friction points.
Using this framework, here are three takeaways on where agentic commerce actually stands today:
-
The Discovery Dilemma is the New SEO
McKinsey notes that for merchants, the "imperative of being found" is the new threshold for dilemma. I couldn't agree more. At VGS, we are still spending the vast majority of our time with clients here today. In an agent-driven world, if your product data isn't structured for an LLM to digest, you don't just rank lower, you cease to exist. The battle for discovery has shifted from "eyeballs on a screen" to "data points in a latent space."
-
The B2B Spectrum: It's Not Just About Enterprise
The discussion around B2B buying often leans heavily toward high-ticket, enterprise sales. However, we see a massive opportunity in the SMB and Solopreneur space.
The automation curve isn't just about the size of the company; it's about the "regret level" of the purchase. A solopreneur buying high-frequency, low-regret supplies is actually a much better candidate for Level 3 "Basic Execution" agents than a massive enterprise with a 6-month procurement cycle. We need to stop viewing B2B as a monolith and start viewing it through the lens of purchase complexity.
-
The Leap to Level 4: The Data Access Divide
McKinsey outlines a path to Level 4 (Autonomization), but I believe we are hitting a significant wall: Agent Data Access.
Right now:
- Consumers have moved past Level 1 (Assist) and are dabbling in Level 2 (Assembly).
- Agents are technically capable of Level 3 (Basic Execution).
- Merchants are the most spread out; while a few elites are at Level 3, most are still finding their footing.
The barrier to true autonomy (Level 4) is that even the best agents don't yet have enough visibility into data to accurately predict multi-purchase intent for a consumer business profile. Even the best agents don't have enough visibility into purchase history, preferences, and consumption to step in.
This is exactly why the industry is watching OpenClaw so closely. The "Massive Divide" right now isn't tech capability, it's delegation. Will users and businesses trust agents with the keys to their data and their wallets?
The Bottom Line
I expect this shift to happen in B2B before it hits the general consumer market. The incentives for efficiency are higher, and the data pools are often cleaner.
The technology is ready for Level 3. The question is: Are our data architectures and trust models ready to let the agents take the wheel?
