Maarten de BruinThe Acceleration Gap
Numbers, benchmarks and research

The Gap Index

Regulation, Benchmarks and Research. The acceleration gap in numbers.

Where the evidence stands

Reviewed 5 September 2026
18 points
Distance between actual and target office utilization, down from 25 points in 2024 and 2025
JLL benchmark 2026
56%
Global office utilization, against a target most portfolios still do not reach
JLL benchmark 2026
8%
Of organizations have moved past AI pilots into active optimization or scaling
JLL benchmark 2026
95%
Of enterprise AI pilots produced no measurable financial result, on workflow grounds
MIT Project NANDA

REGULATION

EPBD transposition deadline reached across the European Union

The recast Energy Performance of Buildings Directive entered into force on 28 May 2024 and member states had until 29 May 2026 to transpose it into national law. It requires minimum energy performance standards that trigger renovation of the worst-performing 16 percent of non-residential stock by 2030, rising to 26 percent by 2033. New buildings owned by public bodies must be zero-emission from 1 January 2028, and all new buildings from 1 January 2030. The threshold for mandatory building automation and control systems in non-residential buildings drops from 290 kW to 70 kW by 31 December 2029. Life-cycle global warming potential must be calculated and disclosed for new buildings above 1,000 square meters from 1 January 2028, and for all new buildings from 1 January 2030.

Key takeaway

The automation threshold is the line that matters most for this argument. Dropping it to 70 kW by the end of 2029 pulls a large tier of mid-sized buildings into continuous monitoring for the first time. Those organizations are about to acquire the visibility that larger portfolios have had for a decade, and they will acquire it without the governance to act on it. The gap is not closing. It is being distributed to a wider population of buildings.

Source: European Commission, Energy Performance of Buildings Directive ↗  ·  EUBAC, EPBD transposition and BACS requirements ↗


BENCHMARK

Occupancy benchmark narrows the actual-to-target gap for the first time

JLL's Global Occupancy Planning Benchmark Report 2026, covering 84 organizations and 716 million square feet across six regions, put global office utilization at 56 percent, up from 54 percent in 2025. The gap between actual and target utilization narrowed from 25 percentage points in both 2024 and 2025 to 18 points in 2026. Around 70 percent of employees now work in the office three to five days a week and 62 percent of organizations require fixed in-office days, up from 49 percent. Only 8 percent of organizations have progressed beyond AI pilots to active optimization or scaling, while more than 70 percent have not begun. Technical spaces make up 23 percent of portfolios and run at 45 percent utilization against a 72 percent target.

Key takeaway

Two numbers in the same report tell the whole story. The distance between measured and intended utilization is finally closing, which means the evidence is being acted on somewhere. And 8 percent of organizations have moved past pilots, which means it is being acted on in very few places. A 27-point shortfall in technical spaces, close to a quarter of the portfolio, is exactly the kind of stable, expensive, cross-domain finding this book is about. It has been visible long enough to be benchmarked.

Source: JLL, Global Occupancy Planning Benchmark Report 2026 ↗


DISCLOSURE

Omnibus I narrows CSRD scope and pushes first reports to 2028

The Omnibus I directive, formally Directive (EU) 2026/470, was published in the Official Journal on 26 February 2026 and entered into force on 18 March 2026. Corporate sustainability reporting now applies to EU companies exceeding both 1,000 employees and 450 million euro in net annual turnover. In-scope EU companies report for financial years starting on or after 1 January 2027, with first reports due in 2028. Non-EU groups follow a year later, reporting for financial years from 1 January 2028 with first reports in 2029.

Key takeaway

A narrower scope and a later date is being read in many organizations as breathing room. That reading is a mistake for anyone holding a large estate. The reporting obligation was never the hard part. Building a governance structure that can act on what the reporting reveals takes years, and the organizations now standing down their programs are giving back exactly the preparation time the delay handed them. The question a disclosure regime eventually forces is not what did you know. It is how long you knew it before you did anything.

Source: Latham & Watkins, conclusion of the sustainability omnibus process ↗  ·  EUR-Lex, Directive (EU) 2026/470 ↗


BENCHMARK

Utilization climbs to 53 percent while data quality becomes the stated obstacle

CBRE's 2026 Global Workplace and Occupancy Insights, drawn from 303 million square feet across portfolios averaging 5 million square feet, reported office utilization at 53 percent against 38 percent in 2024 and 35 percent in 2023. Collaboration was named the most important reason to come in by 68 percent of respondents. Optimizing the real estate portfolio was the primary objective for 80 percent of corporate real estate teams, and 55 percent identified data quality problems and insufficient expertise as the main obstacles to deploying AI and analytics.

Key takeaway

80 percent name portfolio optimization as the primary objective and 55 percent name data quality as the obstacle standing in the way. Those two figures sit next to each other in the same survey and they do not fit together comfortably. Portfolios reporting measured utilization at 53 percent, against 35 percent three years ago, are not short of occupancy evidence. Naming data quality as the constraint is how a governance problem gets budgeted as a technology problem, and it is the most common single misdiagnosis in this field.

Source: CBRE, 2026 Global Workplace & Occupancy Insights ↗


RESEARCH

MIT finds 95 percent of enterprise AI pilots produce no measurable return

Research from MIT's Project NANDA examining enterprise generative AI deployments found that around 95 percent of pilots delivered no measurable effect on profit and loss. The reported cause was not model quality. It was a learning gap: the systems did not retain feedback or adapt to context, and the pilots did not connect to the processes where decisions are actually made. The study drew on 52 structured interviews, 153 survey responses and more than 300 publicly disclosed deployments, and its headline figure has been challenged on methodological grounds.

Key takeaway

This is the acceleration gap arriving in a new domain and behaving identically. A capability enters the organization, produces demonstrable output, and fails to reach the decision. Facility and real estate teams now buying AI into their analytics stack should read this as the warning it is. An agent that produces a better finding faster, delivered into the same governance structure that has been absorbing findings for years, will be absorbed at higher resolution and greater speed.

Source: MIT NANDA, The GenAI Divide: State of AI in Business 2025 ↗