Why companies are doing this
Puestario uses a written test to judge each internal AI agent job in Doral and Miami. The research below explains why repeat work and information chasing deserve attention. It does not promise your business the same results.
Third party figures, each drawn from its source and linked to it. None is a Puestario result.
Son Hábitos operating case
See how we use Emily inside Son Hábitos
Follow real workflows from our own business: voice requests, record lookups, reports and sales handoffs. See the original overview and explore visual examples.
See the agents at work ↗Each study stands on its own. The time shares overlap and must not be added together. None measures a Puestario desk.
2025 · Time use
When work gets in the way of work: Reclaiming organizational capacity
The share of their day that business and HR leaders say goes to work that does not add to what their organization creates.
Study scope: a survey of nearly 10,000 business and HR leaders in 93 countries.
What this study cannot tell you: Self-reported time use. It does not tell us how much work an agent can take over.
Deloitte, 2025 Global Human Capital Trends2025 · Time use
Deloitte's 2025 Global Human Capital Trends press release
Nearly the share of a manager's time spent solving today's problems and on administrative tasks, against 13 percent spent developing their people.
Study scope: the same survey.
What this study cannot tell you: A survey of managers’ time. These categories are not a single measured budget for your business.
Deloitte, 2025 Global Human Capital Trends2021 · Time use
Managers' work and behaviour patterns in profitable growth SMEs
The share of working time that 133 managers of profitable, growing small and mid sized firms (average 21 employees) reported spending on deskwork, inside a 52.4 hour week.
Study scope: a standardized questionnaire answered by 133 managers of Swedish small and mid sized firms.
What this study cannot tell you: Swedish growing firms, not Florida tire shops. Deskwork is not the same thing as automatable work.
Small Business Economics, Ahmadi, Macassa, and Larsson2026 · Adoption
AI use by businesses, collection period ending May 3, 2026
Fewer than one in five firms with four or fewer employees reported using AI, against 37 percent of firms with 250 or more. The gap is the small business.
Study scope: a Census Bureau survey of U.S. businesses on whether they use AI.
What this study cannot tell you: Reported adoption measures use, not business results or value.
U.S. Census Bureau, Business Trends and Outlook Survey2025 · Task performance
Generative AI at Work
The average rise in issues resolved per hour when 5,172 customer support agents got an AI assistant. Less experienced workers gained the most.
Study scope: a peer reviewed study of support agents at one software company.
What this study cannot tell you: Customer support at one software company. This is not a Puestario result or a forecast for your job.
The Quarterly Journal of Economics, Brynjolfsson, Li, and Raymond2023 · Task performance
Experimental evidence on the productivity effects of generative artificial intelligence
The drop in time taken on occupation specific writing tasks for 453 professionals given an AI assistant, with output quality rated 18 percent higher.
Study scope: a preregistered experiment on short writing tasks.
What this study cannot tell you: Short professional writing tasks. This does not establish performance on an ongoing operational job.
Science, Noy and Zhang2026 · Market context
Census Bureau resources for small businesses, Small Business Week
U.S. firms with at least one employee but fewer than 500, in 2023. The businesses this site is written for are among them.
Study scope: a count of employer firms.
What this study cannot tell you: A count of employer firms. Market size is not evidence that an AI desk will work.
U.S. Census BureauOur commitment
What we actually
put in writing.
Your Job Card defines success for your business. The published studies do not replace that test.
- The Job CardOne page. The job, where the machine goes, the pass condition, the fail condition, the install date, and the review date. Signed by both sides before install.
- When the test is checkedAfter installation, on the review date written on the Job Card. We check the result against the written test together.
- The finish lineWritten on the Job Card before anything is installed, so both sides know what done means.
- Corrections and supportWe review the agreed results with your team, document any gaps, and plan corrections within the signed scope. Covered setup faults are handled through the included support.
Ideas and direction
Where AI leaders see this going
Original sources on personal agents and useful AI. These are industry perspectives, not endorsements of Puestario or promises of your results.
Garry Tan: Own Your Intelligence
August 2026. Personal context, owned memory, and reusable procedures. His “personal AGI” framing is not a certification of general intelligence.
Mark Zuckerberg: Personal Superintelligence
July 2025. A vision of AI helping people pursue their goals; a vision is different from a measured result.
Demis Hassabis: a universal AI assistant
May 2025. Context, planning, and action across devices. Research goals and released features should be distinguished.