Simon Roberts (@digitalcampaign) header
Simon Roberts (@digitalcampaign) profile picture

Just a digital creative, developer and solution architect who loves innovation and elegant design-lead, digital solutions.

Nottinghamshire
Joined January 2009
Photo by Simon Roberts (@digitalcampaign)
Most firms treat AI agents like features. In reality, they’re delegated operating capacity. If you can’t name the owner, authority and exit path, you don’t have governance — you have hidden risk. digital-campaign.com/posts/2…
16
Most AI governance fails in the moment that matters: while the system is acting. Static policies can’t stop a bad tool call or customer message in real time. That’s why runtime governance matters. digital-campaign.com/posts/2…
20
The real shift isn’t AI browsing the web. It’s websites exposing a second, machine-readable interface for agents too. That changes design from usability only to usability plus access. digital-campaign.com/posts/2…
1
36
MCP’s real shift isn’t AI in the stack. It’s AI becoming the interface: discoverable capabilities, governed actions, less app-hopping. That changes how GTM software is bought and used. digital-campaign.com/posts/2…
17
Most firms don’t have a software problem. They have a context problem: work is split across too many systems for people or AI to act cleanly. In AI-native workspaces, the prize is orchestration, not consolidation. digital-campaign.com/posts/2…
20
Most AI governance still focuses on models. The real control point is the agent: once software can act, identity decides what it may do, in what context, and who can be held accountable. digital-campaign.com/posts/2…
24
The real prize isn’t an AI gateway. It’s the control layer that decides which model runs, at what cost and with what fallback. That’s where AI unit economics get managed, not just measured. digital-campaign.com/posts/2…
17
The AI race is moving up a layer. The model is no longer the product; the runtime around it is. Tools, memory, permissions and orchestration now decide whether an agent is useful, safe and cost-effective. digital-campaign.com/posts/2…
13
The real shift isn’t GUI to chat. It’s software becoming discoverable to machines at runtime, so agents can choose the right capability on the fly. That changes how products are built, secured and priced. digital-campaign.com/posts/2…
1
20
Human approval is a weak safety model for autonomous systems. At scale, it gets fast, shallow and unreliable. Real control comes from bounded authority, hard limits and intervention where it matters most. digital-campaign.com/posts/2…
20
The real shift isn’t “trusting AI”. It’s making trust an execution control: who acted, under what authority, with what evidence, at the moment of action. That’s how enterprises stay accountable at scale. digital-campaign.com/posts/2…
27
Most firms are watching AI outputs and missing the real risk: agents that can act. Once systems can change records or trigger workflows, control is an identity and permissions problem, not a model one. digital-campaign.com/posts/2…
30