Our predictions.
Five countdowns.

We reassessed all five forecasts against current primary sources on 2 November. New monitoring, enterprise-workflow and software-delivery reports sharpen the evidence, but none meets a forecast's resolution criteria. Robotics and education evidence remains materially unchanged. Every target, milestone and criterion stays unchanged, and all nine earlier entries are preserved.

Change at this review: No days added or removed. Passing days are not included in this change.

Our prediction: AI agents will run repeatable business workflows across at least ten organisations, with people managing exceptions.

Deadline: 31 Dec 2028

What would count

1. Customer-confirmed or independently evaluated public reports name ten organisations across at least three sectors. Each runs two distinct production workflows spanning at least two business systems.
2. Each workflow reports at least 1,000 consecutive eligible cases over at least 90 days. At least 80% finish without a person executing workflow steps. Eligibility, acceptance tests and escalation rules are fixed before the reporting period.
3. People may authorise consequential decisions and handle exceptions. Reports include failures and escalations in the denominator. This predicts repeatable operation at ten organisations, not majority adoption or unattended businesses.

Why this estimate: We retain end-2028. The ServiceNow integration strengthens the case that agents will act inside governed business systems, and the Uber rollout shows substantial organisational commitment. Neither source measures repeatable autonomous completion. We found no customer-confirmed 90-day evidence meeting the two-workflow, ten-organisation and consecutive-case thresholds.

Uncertainty: Timing remains uncertain. Vendor case studies rarely publish consecutive-case failure and intervention data, so the milestone could occur before enough public evidence exists to verify it.

Earlier: Customer-confirmed 90-day results show high autonomous completion across multiple systems and sectors, including failures and recovery costs.

Later: Manual rescue, fragile integrations or operating costs prevent teams expanding beyond isolated queues.

We reassessed all five forecasts against primary sources on 26 October. New agent-control findings, a company-wide workplace rollout, planned multi-site robot deployment and an AI-assisted vulnerability response sharpen the leading indicators and evidence limits. No source meets a forecast's resolution criteria, so every target, milestone and criterion remains unchanged. All eight earlier entries are preserved.

Change at this review: No days added or removed. Passing days are not included in this change.

Our prediction: AI agents will run repeatable business workflows across at least ten organisations, with people managing exceptions.

Deadline: 31 Dec 2028

What would count

1. Customer-confirmed or independently evaluated public reports name ten organisations across at least three sectors. Each runs two distinct production workflows spanning at least two business systems.
2. Each workflow reports at least 1,000 consecutive eligible cases over at least 90 days. At least 80% finish without a person executing workflow steps. Eligibility, acceptance tests and escalation rules are fixed before the reporting period.
3. People may authorise consequential decisions and handle exceptions. Reports include failures and escalations in the denominator. This predicts repeatable operation at ten organisations, not majority adoption or unattended businesses.

Why this estimate: We retain end-2028. The Uber rollout is a substantial access commitment and makes broad workplace use more plausible. It does not show that repeatable workflows finish reliably without people executing steps. The long-running research account still lacks accepted-outcome denominators. We found no customer-confirmed 90-day evidence that meets the declared ten-organisation test.

Uncertainty: Timing remains uncertain. Vendor case studies rarely publish consecutive-case failure and intervention data, so the milestone could occur before enough public evidence exists to verify it.

Earlier: Customer-confirmed 90-day results show high autonomous completion across multiple systems and sectors, including failures and recovery costs.

Later: Manual rescue, fragile integrations or operating costs prevent teams expanding beyond isolated queues.

We reassessed all five forecasts against primary sources on 19 October. New evaluator-access terms, long-running agent accounts, fleet retirement and school tutoring studies sharpen the evidence limits. All targets, milestones and resolution criteria remain unchanged. Earlier entries are preserved.

Change at this review: No days added or removed. Passing days are not included in this change.

Our prediction: AI agents will run repeatable business workflows across at least ten organisations, with people managing exceptions.

Deadline: 31 Dec 2028

What would count

1. Customer-confirmed or independently evaluated public reports name ten organisations across at least three sectors. Each runs two distinct production workflows spanning at least two business systems.
2. Each workflow reports at least 1,000 consecutive eligible cases over at least 90 days. At least 80% finish without a person executing workflow steps. Eligibility, acceptance tests and escalation rules are fixed before the reporting period.
3. People may authorise consequential decisions and handle exceptions. Reports include failures and escalations in the denominator. This predicts repeatable operation at ten organisations, not majority adoption or unattended businesses.

Why this estimate: We retain end-2028. Sustained research sessions and broader product work make useful agent deployment more plausible. The need to inspect unfinished work and escalate also matters. These reports do not establish reliable end-to-end completion across the declared organisations, systems and case cohorts. We found no outcome evidence strong enough to revise the date.

Uncertainty: Timing remains uncertain. Vendor case studies rarely publish consecutive-case failure and intervention data, so the milestone could occur before enough public evidence exists to verify it.

Earlier: Customer-confirmed 90-day results show high autonomous completion across multiple systems and sectors, including failures and recovery costs.

Later: Manual rescue, fragile integrations or operating costs prevent teams expanding beyond isolated queues.

We reassessed all five milestones on 12 October. The new mathematics release, its recorded corrections, CRM previews and connected software-delivery announcements add progress with important verification limits. We reassessed robot operating reports and compared positive short-follow-up tutoring evidence with the negative university trial. None provides a defensible reason to change a deadline or resolves a declared milestone. All five targets and resolution criteria remain unchanged, with zero days added or removed.

Change at this review: No days added or removed. Passing days are not included in this change.

Our prediction: AI agents will run repeatable business workflows across at least ten organisations, with people managing exceptions.

Deadline: 31 Dec 2028

What would count

1. Customer-confirmed or independently evaluated public reports name ten organisations across at least three sectors. Each runs two distinct production workflows spanning at least two business systems.
2. Each workflow reports at least 1,000 consecutive eligible cases over at least 90 days. At least 80% finish without a person executing workflow steps. Eligibility, acceptance tests and escalation rules are fixed before the reporting period.
3. People may authorise consequential decisions and handle exceptions. Reports include failures and escalations in the denominator. This predicts repeatable operation at ten organisations, not majority adoption or unattended businesses.

Why this estimate: We retain end-2028. Cross-system previews make the intended workflow architecture more concrete, and Barclays remains evidence of substantial deployed assistance. The new accounts are too short or incomplete to establish repeatable autonomous completion with exceptions counted. These sources support continued progress without a defensible date adjustment. We still need customer-confirmed denominators and sustained results across sectors.

Uncertainty: Timing remains uncertain. Vendor case studies rarely publish consecutive-case failure and intervention data, so the milestone could occur before enough public evidence exists to verify it.

Earlier: Customer-confirmed 90-day results show high autonomous completion across multiple systems and sectors, including failures and recovery costs.

Later: Manual rescue, fragile integrations or operating costs prevent teams expanding beyond isolated queues.

We reassessed all five milestones against primary material retrieved on 5 October. Fresh evidence shows stronger bounded model, coding and enterprise performance, alongside a large randomised university trial in which access to an AI tutor reduced grades and participation. The robotics evidence adds an adaptable demonstration and an estimated economics model, not the customer operating record our test requires. None of the new material meets a declared resolution rule or justifies changing a target. Every deadline and criterion is retained, with zero days added or removed.

Change at this review: No days added or removed. Passing days are not included in this change.

Our prediction: AI agents will run repeatable business workflows across at least ten organisations, with people managing exceptions.

Deadline: 31 Dec 2028

What would count

1. Customer-confirmed or independently evaluated public reports name ten organisations across at least three sectors. Each runs two distinct production workflows spanning at least two business systems.
2. Each workflow reports at least 1,000 consecutive eligible cases over at least 90 days. At least 80% finish without a person executing workflow steps. Eligibility, acceptance tests and escalation rules are fixed before the reporting period.
3. People may authorise consequential decisions and handle exceptions. Reports include failures and escalations in the denominator. This predicts repeatable operation at ten organisations, not majority adoption or unattended businesses.

Why this estimate: We retain end-2028. Barclays adds strong scale evidence for AI-assisted knowledge retrieval and message routing, while the Agents API announcement shows a broader set of production claims. Neither source distinguishes end-to-end autonomous completion from assistance or publishes the fixed eligibility, 90-day acceptance, failure and rescue accounting required across ten organisations. The operational direction strengthens, but the forecast does not move.

Uncertainty: Timing remains uncertain. Vendor case studies rarely publish consecutive-case failure and intervention data, so the milestone could occur before enough public evidence exists to verify it.

Earlier: Customer-confirmed 90-day results show high autonomous completion across multiple systems and sectors, including failures and recovery costs.

Later: Manual rescue, fragile integrations or operating costs prevent teams expanding beyond isolated queues.

We reassessed all five milestones against primary material retrieved on 28 September. New customer and field evidence strengthens the case for bounded agents, single-task factory robots and AI tutoring at useful scale. It still falls short of our declared denominators, cross-organisation breadth, intervention and economics reporting, production monitoring and durable learning thresholds. We therefore retain every target and resolution rule, with zero days added or removed.

Change at this review: No days added or removed. Passing days are not included in this change.

Our prediction: AI agents will run repeatable business workflows across at least ten organisations, with people managing exceptions.

Deadline: 31 Dec 2028

What would count

1. Customer-confirmed or independently evaluated public reports name ten organisations across at least three sectors. Each runs two distinct production workflows spanning at least two business systems.
2. Each workflow reports at least 1,000 consecutive eligible cases over at least 90 days. At least 80% finish without a person executing workflow steps. Eligibility, acceptance tests and escalation rules are fixed before the reporting period.
3. People may authorise consequential decisions and handle exceptions. Reports include failures and escalations in the denominator. This predicts repeatable operation at ten organisations, not majority adoption or unattended businesses.

Why this estimate: We retain end-2028. Smarsh adds an explicit customer-support denominator and Microsoft describes a sizeable internal portfolio of bounded production agents. Both make the operational trend more credible. Smarsh's 72% deflection is below our threshold and its sample is small, while Microsoft's selected-workflow cycle-time claim is not an autonomous completion rate. Neither supplies the fixed eligibility, failure accounting, 90-day reporting and ten-organisation coverage required to move the date.

Uncertainty: Timing remains uncertain. Vendor case studies rarely publish consecutive-case failure and intervention data, so the milestone could occur before enough public evidence exists to verify it.

Earlier: Customer-confirmed 90-day results show high autonomous completion across multiple systems and sectors, including failures and recovery costs.

Later: Manual rescue, fragile integrations or operating costs prevent teams expanding beyond isolated queues.

We reassessed all five milestones against primary material retrieved on 21 September. Long-horizon formal mathematics, high-volume customer workflows and cross-environment robot generalisation strengthen the evidence that bounded systems are improving. They still do not provide the independent breadth, consecutive eligible-case results, production-ticket outcomes, commercial robot economics or delayed learning replication in our tests. We therefore retain every target and resolution rule, with zero days added or removed.

Change at this review: No days added or removed. Passing days are not included in this change.

Our prediction: AI agents will run repeatable business workflows across at least ten organisations, with people managing exceptions.

Deadline: 31 Dec 2028

What would count

1. Customer-confirmed or independently evaluated public reports name ten organisations across at least three sectors. Each runs two distinct production workflows spanning at least two business systems.
2. Each workflow reports at least 1,000 consecutive eligible cases over at least 90 days. At least 80% finish without a person executing workflow steps. Eligibility, acceptance tests and escalation rules are fixed before the reporting period.
3. People may authorise consequential decisions and handle exceptions. Reports include failures and escalations in the denominator. This predicts repeatable operation at ten organisations, not majority adoption or unattended businesses.

Why this estimate: We retain end-2028. Wayfair's reported monthly volume and Circles' reported autonomous resolution rate are stronger operational signals than activity or response confidence alone. They still lack our predeclared consecutive-case denominator, rescue accounting, 90-day window and ten-organisation coverage. The evidence moves the judgement from isolated examples towards scaled bounded workflows, but not enough to change the deadline.

Uncertainty: Timing remains uncertain. Vendor case studies rarely publish consecutive-case failure and intervention data, so the milestone could occur before enough public evidence exists to verify it.

Earlier: Customer-confirmed 90-day results show high autonomous completion across multiple systems and sectors, including failures and recovery costs.

Later: Manual rescue, fragile integrations or operating costs prevent teams expanding beyond isolated queues.

We reassessed all five milestones against primary material on 18 September. We retain every target and resolution rule. The evidence supports useful bounded capabilities, but does not supply the breadth, consecutive-case denominators, operating economics or delayed learning results our tests require. This is a limited source review, not a claim that all relevant research was surveyed. No date moves because time has passed.

Change at this review: No days added or removed. Passing days are not included in this change.

Our prediction: AI agents will run repeatable business workflows across at least ten organisations, with people managing exceptions.

Deadline: 31 Dec 2028

What would count

1. Customer-confirmed or independently evaluated public reports name ten organisations across at least three sectors. Each runs two distinct production workflows spanning at least two business systems.
2. Each workflow reports at least 1,000 consecutive eligible cases over at least 90 days. At least 80% finish without a person executing workflow steps. Eligibility, acceptance tests and escalation rules are fixed before the reporting period.
3. People may authorise consequential decisions and handle exceptions. Reports include failures and escalations in the denominator. This predicts repeatable operation at ten organisations, not majority adoption or unattended businesses.

Why this estimate: We retain end-2028. Air India's customer-endorsed vendor announcement describes useful multi-system action, but its response-confidence threshold is not an observed autonomous completion rate. Anthropic's tool-call analysis cannot reconstruct complete API sessions. Neither supplies ten organisations' consecutive-case results over 90 days. These gaps do not disprove progress, but they leave insufficient evidence to revise our date.

Uncertainty: Timing remains uncertain. Vendor case studies rarely publish consecutive-case failure and intervention data, so the milestone could occur before enough public evidence exists to verify it.

Earlier: Customer-confirmed 90-day results show high autonomous completion across multiple systems and sectors, including failures and recovery costs.

Later: Manual rescue, fragile integrations or operating costs prevent teams expanding beyond isolated queues.

We reassessed all five predictions against their published tests. The reviewed deployments, evaluations and tutoring studies do not justify moving a target or changing a milestone. All five dates stay unchanged. This is an evidence review, not a claim that another day passing changes our forecast.

Change at this review: No days added or removed. Passing days are not included in this change.

Our prediction: AI agents will run repeatable business workflows across at least ten organisations, with people managing exceptions.

Deadline: 31 Dec 2028

What would count

1. Customer-confirmed or independently evaluated public reports name ten organisations across at least three sectors. Each runs two distinct production workflows spanning at least two business systems.
2. Each workflow reports at least 1,000 consecutive eligible cases over at least 90 days. At least 80% finish without a person executing workflow steps. Eligibility, acceptance tests and escalation rules are fixed before the reporting period.
3. People may authorise consequential decisions and handle exceptions. Reports include failures and escalations in the denominator. This predicts repeatable operation at ten organisations, not majority adoption or unattended businesses.

Why this estimate: We retain end-2028. Air India's expanded workflows strengthen the case that bounded agents can do useful operational work. The announcement still lacks consecutive-case completion and intervention rates across the organisations our test requires. Longer agent activity in Anthropic's study is not equivalent to completed business workflows. We have no sufficient verified basis for moving the date.

Uncertainty: Timing remains uncertain. Vendor case studies rarely publish consecutive-case failure and intervention data, so the milestone could occur before enough public evidence exists to verify it.

Earlier: Customer-confirmed 90-day results show high autonomous completion across multiple systems and sectors, including failures and recovery costs.

Later: Manual rescue, fragile integrations or operating costs prevent teams expanding beyond isolated queues.

We are publishing our own forecasts with explicit milestones and year-end deadlines. Earlier countdowns followed illustrative scenario windows. Those remain in the alternative-scenario record, not as earlier versions of these predictions. The dates below are our editorial judgements after reviewing these sources, not dates supplied by the sources.

Our prediction: AI agents will run repeatable business workflows across at least ten organisations, with people managing exceptions.

Deadline: 31 Dec 2028

What would count

1. Customer-confirmed or independently evaluated public reports name ten organisations across at least three sectors. Each runs two distinct production workflows spanning at least two business systems.
2. Each workflow reports at least 1,000 consecutive eligible cases over at least 90 days. At least 80% finish without a person executing workflow steps. Eligibility, acceptance tests and escalation rules are fixed before the reporting period.
3. People may authorise consequential decisions and handle exceptions. Reports include failures and escalations in the denominator. This predicts repeatable operation at ten organisations, not majority adoption or unattended businesses.

Why this estimate: Bounded workflow execution already exists. We expect the next two years to turn isolated deployments into repeatable operations across sectors, with permissions, recovery and ownership becoming part of the product. End-2028 is our forecast for that operational milestone.

Uncertainty: Timing remains uncertain. Vendor case studies rarely publish consecutive-case failure and intervention data, so the milestone could occur before enough public evidence exists to verify it.

Earlier: Customer-confirmed 90-day results show high autonomous completion across multiple systems and sectors, including failures and recovery costs.

Later: Manual rescue, fragile integrations or operating costs prevent teams expanding beyond isolated queues.