Meet the Jobs Bot
Augmentive Ventures
Build a bigger future. Together.
More human possibility. More machine capability.
Augmentivity is TVV’s premise and research agenda. The evidence is early and conditional. Model figures are illustrative, not forecasts.
Augmentive Ventures is TVV’s vision for profitable ventures that grow human and machine intelligence together. Our aim is more useful work, more creative capacity and more choice over time—not a business model that depends on mass layoffs.1
Evidence that assistance can improve work
Explore the premises
Beyond automation alone
Automate tasks. Expand what people can do.
Replacing a task is not the same as expanding a future. We choose augmentation-first deployment: use gains to build new ventures, develop skills and create room for people to do more valuable work.3
Intelligence, in both directions
People improve the tools. The tools expand people’s reach.
People bring purpose, judgment, original questions and feedback to improve machine systems. Machines can help people explore, make, test and learn—the TechVest idea: technology vested in a person’s capability, not a person diminished by it.2
Six premises. One expanding horizon.
Not a fixed split. A larger field of possibility.
Our premises are that both partners can improve; a growing total can sustain a larger human contribution even as its share falls; and time released should create real choices. We will test whether creativity can grow without losing diversity, validated feedback can improve the next cycle, and energy per useful result can fall without hiding total demand.467
Creativity can be developed—with limits · Chip efficiency is not energy per useful result · Efficiency does not settle total electricity demand
- Develop both partners
- Grow the total, not a fixed split
- Make time a choice
- Grow creativity; protect diversity
- Learn from validated results
- Count energy per result—and in total
Hardware specification efficiency has improved; that does not establish falling energy per useful result. IEA projects rising total data-centre electricity demand, not falling demand.
Full claims and qualifications
Selected chip specification efficiency improved; this is not metered energy per useful output.6
Rated hardware performance per watt is a different boundary from energy per accepted useful result.
IEA projects increasing electricity use by all data centres.7
Projection, not observation; includes non-AI data-centre demand.
TVV commitments
Design for shared gains. Be accountable for the trade-offs.
We commit to augmentation-first deployment and to planning reinvestment in people, skills and new ventures rather than making headcount reduction the default measure of success. We will include affected people in decisions about saved time, preserve human accountability and report trade-offs as well as gains. These are operating commitments, not a guarantee of no layoffs or a forecast for every employer.3
Why productivity and human outcomes need separate measures
- Augmentation first
- Reinvest in people and ventures
- Give people a voice in time use
- Measure honestly; retain accountability
Promising evidence. No shortcuts.
Useful gains are real. Universal synergy is not established.
Research gives us reasons to build—and reasons to test carefully. These findings concern particular tasks and populations, not a universal productivity multiplier.2
Read the human–AI meta-analysis
Primary evidence
15% more issues resolved per hour
AI-assisted support agents improved, especially less-experienced workers. This was task productivity—not observed shorter workweeks.1
Help is not automatic synergy
Human+AI beat humans alone (g=0.64), but not the better solo partner on average (g=−0.23). These are standardized effects, not percentages.2
Time saved is not yet time off
Danish work-chatbot ever-users reported about 3% of work hours saved; earnings and recorded hours did not measurably change in the study window.3
Creativity can improve. Diversity needs care.
Creativity-enhancement methods averaged g=0.53, mostly on short-term lab tests. Separately, AI story ideas improved ratings while making stories more alike.45
What we don’t know yet: whether two-way improvement produces durable creative growth, sustained acceleration or broadly shared gains in income and free time.
Full claims and qualifications
AI assistance increased support agents’ issues resolved per hour by 15%, with the greatest gains among less-experienced workers.1
Task productivity, not a measured reduction in working hours.
Across 106 experiments, human+AI beat humans alone on average (g=0.64), but underperformed the better solo partner (g=−0.23).2
Hedges’ g is a standardized effect, not a percentage; settings varied widely.
In Denmark, work-chatbot ever-users reported time savings averaging about 3% of work hours; administrative records showed no measurable change in earnings or recorded hours in the study window.3
March 2026 working-paper revision; not evidence of realized leisure.
Adult creativity-enhancement methods showed an average positive effect (g=0.53), mostly on short-term laboratory tests.4
Not evidence of lasting real-world invention or an annual growth rate.
In a short-story experiment, AI ideas improved rated novelty and usefulness, while assisted stories became more alike.5
Task-specific finding; individual ratings and portfolio diversity are separate outcomes.
The augmentive loop
An idea is a beginning. A validated result moves us forward.
People and machines generate ideas, turn selected ideas into implementations, and test them against agreed standards. Validated feedback informs the next round; we will test whether that round genuinely improves after counting review, failures and resources.2
Why the joint system must earn its advantage
People guide machines; machines assist people. Both contribute to implementation, independent validation and feedback into the next cycle.
Proposed mechanism to test; not an established law
From premise to practice
Proposed agenda—not a claim of launched programs or completed pilots
Build ventures. Run pilots. Learn what holds up.
Our proposed venture-studio agenda is to develop useful products and run bounded human–machine pilots with real customers and teams. Compare people alone, machines alone and the partnership; keep the quality bar fixed, include review costs, and redesign when the partnership does not earn its place.2
Measure what makes the future better
More than speed. More than headcount.
We will track complementarity gain, quality-adjusted output per human hour, time released and where it actually goes, creativity alongside diversity, and energy per accepted result alongside total energy. We will also track useful customer value and unit economics, so human benefit and profitable growth are tested together.27
Compare the partnership with the stronger solo baseline · Track the total, not just efficiency
Complementarity gain
Joint performance versus the better solo partner on the same task; show resource budgets and coordination costs.
Quality-adjusted output per human hour
Accepted useful output under a fixed rubric divided by all human effort, including setup, review and correction.
Time released and where it goes
Equal-quality, fixed-output time savings; separately log learning, new work and actual discretionary time.
Creativity plus diversity
Novelty, usefulness and distinct approaches; test persistence and real implementation.
Energy per result and total
Energy per accepted result and total energy at a stated boundary, including failed attempts and overhead.
Customer value and unit economics
Useful outcomes for customers and revenue relative to full delivery costs; not a forecast of profitability.
Help build the bigger future
Bring a hard problem. Help shape a better answer.
We’re looking for founders, researchers, builders, investors and operating teams who want to test two-way augmentivity in the real world. Bring an unmet need, a pilot setting or a way to measure what matters—and help build ventures where stronger machines expand human possibility.2
Start with the evidence, not a guarantee
Build the intelligence. Grow the possibility. Keep people in the future.
Sources
- Brynjolfsson, Li & Raymond — Generative AI at Work, QJE (2025)Published field study
- Vaccaro, Almaatouq & Malone — Human–AI combinations, Nature Human Behaviour (2024)Published systematic review and meta-analysis
- Humlum & Vestergaard — Still Waters, Rapid Currents, NBER 33777 (March 2026 revision)Working paper; survey and administrative records
- Haase, Hanel & Gronau — Creativity enhancement methods for adults (online 2023; issue 2025)Published meta-analysisAlternate URL
- Doshi & Hauser — Generative AI, individual creativity and collective diversity, Science Advances (2024)Published short-story experiment
- Epoch AI / Rahman — ML hardware energy efficiency (2024)Hardware specification analysis, not useful-result energy measurement
- IEA — Energy and AI, executive summary (2025)Estimates and Base Case projections; all data centres