Why we build an index.
An index is an indicator you can verify. It forces you to state what counts before the fact, and to be measured against that afterwards. That is why we build one.
We have done this before. In 1999, Reto Ringger launched the first global sustainability index together with Dow Jones, the Dow Jones Sustainability Index ↗. That experience underpins everything that follows. An index is not an exact science. It is a disclosed rule that turns a conviction into a verifiable number and prompts companies to align with it. That once changed the reporting of entire industries, long before it became a requirement.
"Back then, the Dow Jones Sustainability opened up new perspectives for companies and influenced their behaviour. We believe the Globalance Intelligence 100 is capable of doing the same."
Reto Ringger
The core idea
1 the core idea
Not the hundred largest companies. The hundred most compelling.
We evaluate around one hundred companies globally that are actively shaping this phase of the economy. We review them regularly against a consistent set of criteria.
Software, agents, robotics, diagnostics. Without AI, their offering would not exist in this form. They are the bridge: computing power becomes a product for which Business pay. This is where winners change most quickly.
Adopters
Companies across all industries that use AI to achieve more with less labour. This is the large, broad and lasting story. It is missing from pure AI funds. Their effect shows in margin.
| Role | Description | Number of holdings |
| Enablers | the builders | 34 |
| Innovator | the providers | 20 |
| Adopters | the users | 43 |
Model holdings list GI 100, role assignments as of 23 September 2026. The proportions of the three roles are disclosed as observed figures, not managed to a fixed target.
Why now. The Globalance Intelligence Index measures the transition across seven dimensions, each rated on a scale of 0 to 100 against the 2040 target. The result is a gap. What machines can do has advanced considerably. How far that progress has permeated labour markets, politics and society is another matter entirely. That gap is the investment window.
| Dimension | Value |
| AI capability | 66 |
| Geopolitics of AI | 38 |
| Path Divergence | 38 |
| Capital Concentration | 31 |
| Social Adaptation | 19 |
| Labour-market substitution | 19 |
| Policy response | 18 |
The four technology- and capital-proximate dimensions are shaded green, the three societal dimensions in grey/blue.
Index dimension scores, snapshot July 2026 · Scale 0–100 against the 2040 target path. Current values are publicly available under Understand › Index.
The methodology
2 the methodology
Four criteria decide, and each one rules something out.
Every criterion is tied to a number and is reassessed every quarter. A company that breaches a criterion drops out, regardless of how compelling the story sounds.
| Criterion | What we measure | What it excludes |
| Documented exposure | A documented, quantifiable share of the business is tied to the AI transformation. | The well-known company that "also does AI" without it showing up in the numbers. |
| Impact over announcements | For adopters, additionally, a substantiated improvement in margins or efficiency. | The press release with no impact on the profit and loss account. |
| Position in the value chain | Where the company sits in the value chain and whether a bottleneck exists there. | Interchangeable suppliers with no pricing power. |
| AI Footprint | How a company handles the technology itself: oversight, disclosure, treatment of affected stakeholders, security — and which of these is substantiated by evidence. | Companies deploying AI without discernible oversight, and those with documented violations. |
The complete version of these rules is set out in the GI 100 Rulebook. Criteria, thresholds, weighting, rebalancing and exclusions at a glance. Download rulebook (PDF) ↗ · Version 1.1, not for distribution to third parties.
Weighting
Not by size. By conviction and volatility, with hard limits so that conviction does not become concentration risk.
How the index is built
3 how the index is constructed
Five steps, four times a year. The same process, every quarter.
| Step | Number | Label | Note |
| 1 | 37 | leading indicators | with an open source |
| 2 | 3 | Scenarios | possible worlds 2040 |
| 3 | 14 | Investment Themes | where value is created |
| 4 | ~100 | Title | Enablers, Innovators, Adopters |
Based on the concept paper on the investment process of the GI 100, version of 4 August 2026.
Every step is disclosed. We show not only what is in the index, but also why, and what would cause a name to be dropped from it.
What "productive" actually means
4 what "productive" actually means
A Business can buy AI and still not become more productive. Announcements are therefore not a valid selection criterion. For users, the Adopter Score replaces the criterion of 'AI relevance'. It rests on five signals that can be verified from published figures — by us and by anyone who wishes to check our work.
SIGNAL 1
Productivity per head
How we measure this ›
Measured by: Revenue and operating profit per full-time equivalent, assessed on both an absolute basis and in terms of trend over three years. Why: When AI works, the same workforce achieves more, or a smaller workforce achieves the same. Both show up here first. Boundary: Headcount reduction without a gain in output raises the metric too. It is therefore never read in isolation, but always alongside signals 2 and 3.
SIGNAL 2
Margin without headcount growth
How we measure this ›
Measured by: The trend in operating margin alongside stable or declining staff costs as a percentage of revenue. Why: This is the clearest evidence that productivity, rather than price increases, is sustaining the margin. Boundary: In cyclical sectors, a strong sales market can produce the same picture. We therefore always compare against the sector, never in absolute terms.
SIGNAL 3
Sector Efficiency
How we measure this ›
Measured by: The metric by which each sector measures its own efficiency. Cost per transaction in retail, loss ratio in insurance, throughput per plant in manufacturing, processing time per case in banking. Why: A one-size-fits-all metric from outside the industry measures the same thing everywhere and the right thing nowhere. Boundary: This figure differs by sector and is therefore not comparable across sectors. It counts only within its own sector.
SIGNAL 4
Disclosed AI investments
How we measure this ›
Measured by: Confirmed programmes, roll-outs and amounts drawn from annual reports, quarterly releases or investor presentations, with source and date. Why: Those who invest make a commitment. Those who only announce do not. Boundary: This signal alone is never sufficient for inclusion. It supports the others; it does not replace them.
SIGNAL 5
Competitive Impact
How we measure this ›
Measured by: Market share and pricing power — in other words, whether the efficiency advantage stays with the company or is passed on to the market as a price reduction. Why: Productivity that is passed on in full does not increase company value. Boundary: This signal moves slowly. It confirms a thesis; it does not discover one.
No signal is based on an announcement. That is the difference between a thesis and a narrative.
Schematic illustration. The largest drawdown of a pure enabler basket was around 35 per cent in 2022. Model calculation using proxy ETFs.
Schematic illustration of the core problem. A pure builder basket depends on a single factor. The largest drawdown was around 35 per cent in 2022. Model calculation using proxy ETFs.
Individual securities and metrics mentioned illustrate the methodology and do not constitute investment advice. Investments in equities carry risks, including the loss of capital invested.
