Company reportedLive since 2023Added Oct 2026

Repsol

Repsol's Copilot pilot: over two hours saved a week per person, 15 minutes counted as net productivity

Spanish energy group Repsol piloted Microsoft 365 Copilot with hundreds of employees in late 2023. Repsol reported more than two hours of working time saved per person per week. A news report on the study splits that into 121 gross minutes and 15 minutes counted as net productivity. Repsol then gave all staff the web version of Copilot.

Key figures

Copied exactly from the sources. Copy any figure with its citation.

More than two hours

Working time saved per person per week in the late-2023 Copilot Microsoft 365 pilot, as reported by Repsol1

600

Employees who took part in the late-2023 pilot, per Repsol's press release1

121

gross minutes

Average gross time saved per employee per week in the pilot study, as reported by Merca22

15

minutes

Net productivity saving per person per week in the same study, as reported by Merca2; the rest of the 121 minutes was counted as wellbeing benefit2

What it took

CostNot disclosed
TeamNot disclosed
TimelinePilot study ran from October 2023 to January 20242

The problem

Repsol is a large Spanish energy group. When generative AI took off in 2023, it set up a centre of competence in June 2023 to work out how the technology could make its professionals more efficient. Merca2 reports that Repsol decided to measure how much time and work its employees saved with Copilot.

What they built

Repsol tested Microsoft 365 Copilot, an assistant built into Microsoft's office apps such as Teams, Outlook and Word. It brings together web knowledge and the user's own work data. After the pilot, Repsol gave all employees Copilot for Web, a version for web search, document summaries and image generation in a secure setting. It planned a full Microsoft 365 rollout for a later phase, to grow with take-up.

How they did it

Repsol says the pilot ran at the end of 2023 with 600 employees. Merca2 describes a study of 550 employees across all areas, 300 with Copilot and 250 as a control group, with surveys and interviews before and after. The study ran from October 2023 to January 2024. For the wider rollout Repsol added a training programme through 2024, with learning material, help corners, introductory sessions and a course on writing instructions for AI.

Results

Repsol reports more than two hours of working time saved per person per week. Merca2 reports the study counted 121 gross minutes saved per week but only 15 minutes as net productivity. It says the remaining time brought health and wellbeing benefits that may lift engagement. At the end of the test, 61.9% of licensed employees said they would not want to work without the tool.

What went wrong

No failure is reported. The gap between the gross and net figures is the main warning: most of the time saved was not counted as productivity.

All reported figures

MeasureFigureCitation
Working time saved per person per week in the late-2023 Copilot Microsoft 365 pilot, as reported by RepsolMore than two hours1
Employees who took part in the late-2023 pilot, per Repsol's press release6001
Average gross time saved per employee per week in the pilot study, as reported by Merca2121 gross minutes2
Net productivity saving per person per week in the same study, as reported by Merca2; the rest of the 121 minutes was counted as wellbeing benefit15 minutes2
Share of licensed employees who said at the end of the test they would not want to go back to working without the tool61.9%2
Employees in the pilot study per Merca2: 300 with Copilot access and 250 as a control group5502

Lessons for your project

  1. Run the pilot with a control group, as the study did (300 users, 250 without access). It lets you compare like with like instead of relying on opinions.

  2. Separate gross time saved from net productivity. The study counted 121 minutes saved but only 15 minutes as productivity, so a business case built on the larger figure would overstate the gain.

  3. Measure by area and job type. Time saved varied widely across departments in the study, so target licences where the gain is largest.

  4. Be ready to change the rollout. After the pilot Repsol first gave everyone the web version of Copilot and planned the full Microsoft 365 version for a later phase.

Caveats

  • The headline figure comes from Repsol's own press release. The more detailed figures (121 minutes, 15 minutes, 61.9%, 550 participants) come from a Merca2 article that relays the study, which it says was published by MIT Technology Review. Neither source is an audit.
  • The two sources disagree on pilot size: Repsol says 600 employees, Merca2 says 550 (300 with Copilot, 250 control group).
  • Dates differ in wording. Repsol says the pilot ran at the end of 2023. Merca2 says the study was carried out in October 2023 and ended in January 2024. Only the year is used as the deployed date.
  • The pilot tested Microsoft 365 Copilot. The June 2024 press release says the company-wide rollout was the web version, Copilot for Web, with the Microsoft 365 version to follow later depending on take-up. This case does not report results for the company-wide rollout.
  • Neither source names the underlying language model.

Sources

  1. Repsol extiende la IA generativa a todos sus empleados con Copilot

    Repsol · 2024-06-13 · Press release · repsol.com

Summarized in our own words from the sources above. Spotted an error? See corrections.