Machines are poised to take over more of the decision-making and actual work of every company.
And they will do it faster —- this time for real.
Gartner’s most recent Hype Cycle describes a pragmatic “autonomous business” generation catalyzed by machine customers, AI agents, decision intelligence and programmable money, all transitioning from experiments to production-grade impact.
This message is hearing home as when other analysts (Forrester and McKinsey) set their focus on agentic and automation-first innovation as the next big operating value. And for all the splashy headlines about consumer-facing flash, an MIT study indicates that the early returns are highest in back office and operational workflows — precisely where these four trends are breaking out.
Machines, not customers, are putting in the orders
Machine customers are software or devices that make purchases on behalf of people or companies. Think connected printers reordering toner, industrial sensors setting off spare part purchases, or electric vehicles that pay for their own charging. By Gartner’s calculation, there are already billions of these machine customers loosed in the wild, and the installed base will grow into the tens of billions as deployments expand.
The payoff is serious: 24/7 purchasing with no cart abandonment, less stockouts and better alignment between consumption and replenishment. Examples of this model include programs like Amazon’s Dash Replenishment, smart vending and HVAC-as-a-service. In order to do this safely, leaders must possess machine identity, spend limits, and vendor APIs enabling the automatic negotiation and fulfillment with auditability.
AI agents are evolving from copilots to operators
AI agents deliver more than just chat, completing tasks such as monitoring systems, triaging tickets, composing and sending responses, invoking tools and closing the loop. Gartner is predicting that with companies standardizing policies and getting agents connected to enterprise systems, half of routine business decisions will be handled by agents. Customer service, supply chain exceptions, and closing aspects of finance closing tasks are early examples.
Oversight remains essential. Industry surveys reflect deep-seated concern among IT leaders about data loss and agent autonomy, so successful programs establish guardrails first, from role-based access and policy engines to human-in-the-loop checkpoints before higher-risk actions, and continuous red-teaming. In practice, insurers are automatically authorizing low-dollar refunds with clear boundaries, and B2B support centers are routing and solving often for those common cases from beginning to end, drastically reducing handle time the backlog.
Decision intelligence makes judgment into software
Decision intelligence (DI) views decisions as assets to be reused. It mimics how decisions are created, bridges information and action, and creates feedback loops, so that its results get better over time. Restrepo said he and his team have developed through causal inference, simulation, optimization and even digital twins of the organization — so the system not only predicts, it prescribes and learns.
Gartner framers the scope quite succinctly: digitize and model decisions to close the insight-to-action gap. A retailer can test promotions across regions before spending a dollar; a logistics provider can re-optimize routes as weather or capacity patterns shift; a treasury team can recalibrate cash buffers as risk signals change. The pattern is repeatable: define the decision, encode policies, wire up actuators (what can be changed), measure result deltas to feedback into the loop.
Programmable money automates settlement
Programmable money is value rendered as computer code that can enforce its own terms and conditions — payments that are determined by rules and are scheduled and logic driven. With smart contracts and tokenized assets, businesses can introduce automation around processes such as settlement, escrow, compliance checks, and even microtransactions between devices. Central bank and industry experiments, such as those being led by the Bank for International Settlements, the European Central Bank and the Monetary Authority of Singapore, show how conditional payments and atomic delivery-versus-payment help to avoid reconciliation headaches.
For businesses, this could look like subscriptions that are automatically paused when usage drops, equipment that pays out per cycle or supply contracts that trigger funds once IoT-verified milestones are achieved. The to-do list: work toward alignment with ISO 20022 data standards, establish strong key management, and integrate programmable rails into your various ERP and treasury systems so finance keeps full visibility and control.
How leaders should behave now
Begin with a map of high-volume, rules-based decisions — returns, discounts, replenishment, approvals — and codify the policy behind them. Pilot an agent in a limited permission mode, and then scale, communicating with the core using secure APIs. Develop a machine identity and spend-governance architecture that enables devices to safely transact. And get your payments tower ready for conditional, tokenized settlement and get rid of the manual reconciliation.
The common thread through all four of these trends is mundane yet powerful: fewer handoffs, faster cycles, and cleaner data. The largest gains, as the research from MIT points out, are in work that is already structured. Automate the invisible first, measure by results not effort, compound on results—don’t worry about the hype.