Research & evidence
Total Project Management Framework (TPM)
Research shows both gains from AI and risks in how it is used. Task, workflow and evaluation matter.
Brynjolfsson, Li & Raymond · The Quarterly Journal of Economics
Across 5,172 customer-support agents, AI access increased issues resolved per hour by 15% on average. Effects varied by experience; the study also found evidence of learning.
One firm and one kind of work. These results do not establish the effect of TPM.
Read the primary source ↗Lee et al. · CHI · Microsoft Research / CMU
In a survey of 319 knowledge workers, greater confidence in generative AI was associated with less reported critical-thinking effort.
Self-reports describe perceived effort. They do not show that AI caused lasting loss of skill.
Read the primary source ↗Shen & Tamkin · Anthropic Fellows Program / Anthropic · Research preprint
In a study of learning a new programming library, AI assistance reduced immediate understanding on average. Interaction patterns that retained cognitive engagement were associated with better learning.
A specific learning task and immediate assessment. Interaction patterns were not separately randomized; this is not a test of long-term professional skill.
Read the primary source ↗Dell’Acqua et al. · Organization Science · BCG consultants
In a randomized experiment with 758 BCG consultants, a 373-person group tackled a business problem beyond the tested model’s capabilities. GPT-4 assistance reduced correct decisions, while AI-assisted recommendations were judged more coherent and persuasive—even when wrong.
The experiment used GPT-4 in 2023. Results concern specific tasks, not today’s models, long-term skill loss, or the effectiveness of TPM.
Read the primary source ↗A tool can improve current performance. That alone does not tell us what people learn.
Visible steps, editable work and meaningful checks are plausible ways to support participation.
TPM needs its own evaluation. These studies motivate the question; they do not answer it for TPM.
| Evaluation focus | What to examine |
|---|---|
| Quality of work | Accuracy, useful evidence, omissions and the quality of decisions. |
| Understanding | Whether people can explain assumptions and identify weak reasoning. |
| Transfer | Whether they can apply the method to a new task with less assistance. |
| Time and effort | Where assistance saves effort and where human attention remains necessary. |
Proposed evaluation criteria. No completed TPM study or customer outcome is claimed here.
Not a member yet? Register now
Are you a member? Login now
