From Tactical Tools to Decision Centres: Enterprise AI Today

This is part 1 of an essay series Automated Authority at Work , created as a submission to a course on AI Saftey & Governance.

If you work in a typical enterprise setting, AI might feel like it's everywhere, but its use is likely restricted to enhancing how people get their work done, and finding the information they need. AI as an decision-authority, or a goal orientated adaptive syste, is very limited today in knowledge-work environments. However, for millions of platform and gig workers, algorithmic management is already their daily reality.

Drivers for Uber and delivery workers for companies like Deliveroo are already directed, evaluated, and even terminated by algorithms with minimal human oversight. These workers experience what many professional workers have yet to face: having their livelihood determined by opaque systems they cannot effectively question or appeal.

What we're witnessing isn't the dawn of algorithmic management, but its expansion from casual/gig-platform economies into traditional workplaces. The key question isn't whether machines will take charge of decisions, but how widely this model spreads and what protections will exist when it does.

Trends, vendors and those on the cutting edge tell us that AI can, and will make more and more decisions -- and these decision might be objectively better in a close-probelm environment. But what happens when we get there?

Will we all soon be raging against the machine?

Four emerging technologies show the potential for AI to move to the centre of centralized decision systems that drive an enterprises competitive advantage.

1. Large Action Models (LAMs): Unlike language models that just produce text, LAMs can execute actions directly within business systems. Companies like Orby AI are building models that can manipulate interfaces, trigger workflows, and complete complex multi-step tasks across enterprise applications.

2. Enterprise Intelligence Platforms: Systems like Honu AI's "Decision Infrastructure" are creating cognitive layers that sit between data and operations, centralizing decision-making under algorithmic control with promises of "superhuman decision-making capabilities."

3. Agentic Automation Systems: Fully autonomous AI agents like Manus AI operate as self-sufficient digital workers, independently initiating and executing complex tasks without continuous human direction.

4. Algorithmic Management Systems: As seen in Amazon Supply Chain, these systems make increasingly significant operational decisions - optimizing inventory, determining staffing levels, setting productivity expectations, and routing resources with minimal human oversight.

This shift isn't merely technical advancement - it represents a fundamental redistribution of decision authority within organizations. When AI moves from providing insights to executing decisions autonomously, the power dynamics of the workplace transform.

I don't want to suggest a moral judgment on these systems or this technology -- enterprises have to compete, and these technologies offer real potential to do just that -- none of this is either objectively good or bad -- but it's still valuable to think about where the ball is going...

Warning Signs of Ever-Deeper Two-Tier Workforces

I don't kniow exactly which of these vendors, technical approaches or value propositions will be adopted at scale in the years ahead, all point toward a bifurcation in workplaces that's likely to accelerate:

• Few executives: will find that AI can serve to augment their breadth and depth of control, enhancing their capabilities and extending influence. These employees operate alongside AI, retaining agency over which recommendations to accept or reject.

• Most operational workers: including white-collar workers and middle managers may find that AI increasingly functions as an automated manager, assigning tasks, measuring performance, and making consequential decisions about scheduling, evaluation, and advancement opportunities with limited transparency or appeal mechanisms.

Ironically, these same workers may benefit from AI tools Co-Pilots and generative AI in executing their daily tasks, creating an illusion that they too are beneficiaries of AI advancement. The surface-level empowerment is, in other words masking a deeper reality that they are simultaneously subject to more consequential AI-driven decision systems that could determine their career paths, compensation, and opportunities.

Here's the kicker about awareness: Uber drivers, know the algorithm is their boss. But PM's (like me) excitedly using a new tool might be missing that other tools are silently tracking their output, response times, and collaboration patterns etc.

As The Centre of Data Ethics & Innovation note: "The adoption of algorithms in decision-making poses challenges because such algorithms may reflect or amplify existing power imbalances." This dynamic is precisely what we risk overlooking as AI becomes both a tool and an overseer in knowledge work environments.

Interestingly this week Microsoft launched their latest offering to infer the skills enterprise staff develop through their interactions on Microsoft real estate. This is practical and exciting for colleagues in People Analytics, but hints at an underlying truth that our skills are determined by a machine, and how we work is monitored to such granularity, that the replication of work feels innevitable.

For futher reading on this topic, I highly recommend this "The future of work? Inequality, the advance of Artificial Intelligence, and what can be done about it: A literature review", by Caleb Peppiatt.

The next essay in this series, Delusions of Objectivity: AI has Psychological Leverage On Us All -- will look at the potential for decision making AI to be good, fair and kept in check.