Introducing Tendem:
The first Hybrid AI + Human agent
Tendem is a hybrid AI + Human agent that combines AI automation with human expertise to deliver business-ready results. It automates the entire workflow—from task breakdown to expert matching and quality verification—resulting in higher-quality outputs than AI-only tools and faster, better results than human-only marketplaces.
Tendem performance validation
To assess Tendem's success, we conducted rigorous in-house testing across 94 diverse, hard real-world business tasks. The benchmark measures client-perceived outcomes across three key dimensions: result quality, execution time, and price.
Tendem Internal Benchmark
Benchmark Composition: 94 tasks across four key business functions
74.5%
Tendem
53.2%
Upwork
40.4%
ChatGPT Agent
16.4 hours
Tendem
35.0 hours
Upwork
ChatGPT Agent
Tendem Benchmark Insights
![](data:image/svg+xml,<svg display="block" role="presentation" viewBox="0 0 48 48" xmlns="http://www.w3.org/2000/svg"><path d="M 19.004 38 C 16.376 38 13.905 37.501 11.593 36.504 C 9.281 35.507 7.269 34.153 5.559 32.444 C 3.849 30.734 2.495 28.723 1.496 26.412 C 0.499 24.101 0 21.631 0 19.004 C 0 16.376 0.499 13.905 1.496 11.593 C 2.493 9.281 3.847 7.269 5.556 5.559 C 7.266 3.849 9.277 2.495 11.588 1.496 C 13.899 0.499 16.369 0 18.996 0 C 21.624 0 24.095 0.499 26.407 1.496 C 28.719 2.493 30.731 3.847 32.441 5.556 C 34.151 7.266 35.505 9.277 36.504 11.588 C 37.501 13.899 38 16.369 38 18.996 C 38 21.624 37.501 24.095 36.504 26.407 C 35.507 28.719 34.153 30.731 32.443 32.441 C 30.734 34.151 28.723 35.506 26.412 36.504 C 24.101 37.501 21.631 38 19.004 38 Z M 19 35 C 23.467 35 27.25 33.45 30.35 30.35 C 33.45 27.25 35 23.467 35 19 C 35 14.533 33.45 10.75 30.35 7.65 C 27.25 4.55 23.467 3 19 3 C 14.533 3 10.75 4.55 7.65 7.65 C 4.55 10.75 3 14.533 3 19 C 3 23.467 4.55 27.25 7.65 30.35 C 10.75 33.45 14.533 35 19 35 Z M 19 19 M 18.5 28.5 L 21.5 28.5 L 21.5 9.5 L 14.5 9.5 L 14.5 12.5 L 18.5 12.5 Z" fill="rgb(0, 0, 0)" height="38px" id="jWk0xhs8P" transform="translate(5 5)" width="38px"/></svg>)
Quality First
Tendem improves the overall Good rate by +21.3 percentage points (pp) versus Upwork. The largest quality gain is in Completeness (+22.3 pp), indicating that step-gates prevent omissions and enforce acceptance criteria.
![](data:image/svg+xml,<svg display="block" role="presentation" viewBox="0 0 48 48" xmlns="http://www.w3.org/2000/svg"><path d="M 19.004 38 C 16.376 38 13.905 37.501 11.593 36.504 C 9.281 35.507 7.269 34.153 5.559 32.444 C 3.849 30.734 2.495 28.723 1.496 26.412 C 0.499 24.101 0 21.631 0 19.004 C 0 16.376 0.499 13.905 1.496 11.593 C 2.493 9.281 3.847 7.269 5.556 5.559 C 7.266 3.849 9.277 2.495 11.588 1.496 C 13.899 0.499 16.369 0 18.996 0 C 21.624 0 24.095 0.499 26.407 1.496 C 28.719 2.493 30.731 3.847 32.441 5.556 C 34.151 7.266 35.505 9.277 36.504 11.588 C 37.501 13.899 38 16.369 38 18.996 C 38 21.624 37.501 24.095 36.504 26.407 C 35.507 28.719 34.153 30.731 32.443 32.441 C 30.734 34.151 28.723 35.506 26.412 36.504 C 24.101 37.501 21.631 38 19.004 38 Z M 19 35 C 23.467 35 27.25 33.45 30.35 30.35 C 33.45 27.25 35 23.467 35 19 C 35 14.533 33.45 10.75 30.35 7.65 C 27.25 4.55 23.467 3 19 3 C 14.533 3 10.75 4.55 7.65 7.65 C 4.55 10.75 3 14.533 3 19 C 3 23.467 4.55 27.25 7.65 30.35 C 10.75 33.45 14.533 35 19 35 Z M 19 19 M 13.5 28.5 L 24.5 28.5 L 24.5 25.5 L 16.5 25.5 L 16.5 20.5 L 24.5 20.5 L 24.5 9.5 L 13.5 9.5 L 13.5 12.5 L 21.5 12.5 L 21.5 17.5 L 13.5 17.5 Z" fill="rgb(0, 0, 0)" height="38px" id="BcnBDrDKh" transform="translate(5 5)" width="38px"/></svg>)
Speed
Tendem cuts the median total time by an impressive 53% compared to the human-only baseline. This is driven by faster connection and execution times.
![](data:image/svg+xml,<svg display="block" role="presentation" viewBox="0 0 48 48" xmlns="http://www.w3.org/2000/svg"><path d="M 19.004 38 C 16.376 38 13.905 37.501 11.593 36.504 C 9.281 35.507 7.269 34.153 5.559 32.444 C 3.849 30.734 2.495 28.723 1.496 26.412 C 0.499 24.101 0 21.631 0 19.004 C 0 16.376 0.499 13.905 1.496 11.593 C 2.493 9.281 3.847 7.269 5.556 5.559 C 7.266 3.849 9.277 2.495 11.588 1.496 C 13.899 0.499 16.369 0 18.996 0 C 21.624 0 24.095 0.499 26.407 1.496 C 28.719 2.493 30.731 3.847 32.441 5.556 C 34.151 7.266 35.505 9.277 36.504 11.588 C 37.501 13.899 38 16.369 38 18.996 C 38 21.624 37.501 24.095 36.504 26.407 C 35.507 28.719 34.153 30.731 32.443 32.441 C 30.734 34.151 28.723 35.506 26.412 36.504 C 24.101 37.501 21.631 38 19.004 38 Z M 19 35 C 23.467 35 27.25 33.45 30.35 30.35 C 33.45 27.25 35 23.467 35 19 C 35 14.533 33.45 10.75 30.35 7.65 C 27.25 4.55 23.467 3 19 3 C 14.533 3 10.75 4.55 7.65 7.65 C 4.55 10.75 3 14.533 3 19 C 3 23.467 4.55 27.25 7.65 30.35 C 10.75 33.45 14.533 35 19 35 Z M 19 19 M 13.5 28.5 L 24.5 28.5 L 24.5 20.5 L 23 19 L 24.5 17.5 L 24.5 9.5 L 13.5 9.5 L 13.5 12.5 L 21.5 12.5 L 21.5 17.5 L 17.115 17.5 L 17.115 20.5 L 21.5 20.5 L 21.5 25.5 L 13.5 25.5 Z" fill="rgb(0, 0, 0)" height="38px" id="YBTkOeFGP" transform="translate(5 5)" width="38px"/></svg>)
Closing the Gap
For ambiguous, multi-document, or spec-heavy tasks, the hybrid step-gates substantially cut omission/fabrication errors, solving the "last-mile gap" where AI-only systems fail.
![](data:image/svg+xml,<svg display="block" role="presentation" viewBox="0 0 48 48" xmlns="http://www.w3.org/2000/svg"><path d="M 19.004 38 C 16.376 38 13.905 37.501 11.593 36.504 C 9.281 35.507 7.269 34.153 5.559 32.444 C 3.849 30.734 2.495 28.723 1.496 26.412 C 0.499 24.101 0 21.631 0 19.004 C 0 16.376 0.499 13.905 1.496 11.593 C 2.493 9.281 3.847 7.269 5.556 5.559 C 7.266 3.849 9.277 2.495 11.588 1.496 C 13.899 0.499 16.369 0 18.996 0 C 21.624 0 24.095 0.499 26.407 1.496 C 28.719 2.493 30.731 3.847 32.441 5.556 C 34.151 7.266 35.505 9.277 36.504 11.588 C 37.501 13.899 38 16.369 38 18.996 C 38 21.624 37.501 24.095 36.504 26.407 C 35.507 28.719 34.153 30.731 32.443 32.441 C 30.734 34.151 28.723 35.506 26.412 36.504 C 24.101 37.501 21.631 38 19.004 38 Z M 19 35 C 23.467 35 27.25 33.45 30.35 30.35 C 33.45 27.25 35 23.467 35 19 C 35 14.533 33.45 10.75 30.35 7.65 C 27.25 4.55 23.467 3 19 3 C 14.533 3 10.75 4.55 7.65 7.65 C 4.55 10.75 3 14.533 3 19 C 3 23.467 4.55 27.25 7.65 30.35 C 10.75 33.45 14.533 35 19 35 Z M 19 19 M 21.5 28.5 L 24.5 28.5 L 24.5 9.5 L 21.5 9.5 L 21.5 17.5 L 16.5 17.5 L 16.5 9.5 L 13.5 9.5 L 13.5 20.5 L 21.5 20.5 Z" fill="rgb(0, 0, 0)" height="38px" id="K2WI6uh6i" transform="translate(5 5)" width="38px"/></svg>)
Underlying Strength
Tendem's autonomous AI agent is competitive on web browsing/tool-use and close to leading models on hard knowledge, providing a solid backbone for the hybrid system.
External Industry Benchmarks
Tendem’s AI agent — tested in fully automated mode without any human input — performs on par with leading AI systems on standard industry benchmarks:
System
Humanity’s Last Exam
GAIA
Tendem’s AI Agent
71.0%
39.0%
78.2%
ChatGPT Agent
68.9%
41.6%
—
ChatGPT Deep Research
51.5%
26.6%
67.4%
Manus
—
—
73.4%
Contributors
Toloka AI
Toloka AI