The evidence
“Are soft skills even trainable — and are they really this valuable?”
It’s the question every leader and every professional asks before investing in people-first “power skills.” Fair question. Here is the independent research — randomized controlled trials, labor-market economics, and large-scale workforce studies — that answers it: they are trainable, at any skill level, and they are where the value is moving.
A note on language: you’ll see us call them People First Power Skills. The industry long filed communication, problem-solving, and collaborative leadership under “soft skills” — but their growing priority, and their power over career growth and project outcomes, have earned them a stronger name. We borrow “power skills” from the Project Management Institute, which retitled them in Pulse of the Profession 2023: Power Skills — Redefining Project Success.
Each study below links to the original material, with a note on how it supports the People First Method.
The studies
What the research shows.
258%
net return to the firm within eight months of soft-skills training
Adhvaryu, Kala & Nyshadham — Journal of Political Economy (2023)
A randomized controlled trial across five factories assigned workers at random to on-the-job soft-skills training (communication, time management, problem-solving). Trained workers became ~20% more productive than the control group, and the employer’s net return reached 258% within eight months of completion.
Why it matters here: randomization settles the teachability question — the training caused the gains. And it worked for entry-level workers, proving power skills are learnable at any starting level, with returns that accrue to the organization.
Journal article · NBER working paper · Plain-English summary
+12 pts
growth in social-skill-intensive jobs’ share of the U.S. labor force, 1980–2012
David Deming — Quarterly Journal of Economics (2017)
Harvard economist David Deming found jobs requiring high social interaction grew nearly 12 percentage points as a share of the U.S. labor force, while math-intensive, low-social jobs (including many STEM roles) shrank. Pay grew fastest in roles combining strong technical and social skills — because social skills cut the coordination costs of team production.
Why it matters here: power skills aren’t a substitute for technical skill — they’re the multiplier on it. That pairing is exactly what the People First Method trains.
92%
of talent leaders say soft skills matter as much or more than hard skills — and 89% say bad hires lack them
LinkedIn — Global Talent Trends (2019)
Across thousands of hiring leaders, the consensus is overwhelming — yet the same research shows leaders usually discover a soft-skills gap only after the hire, because these skills are hard to assess from interviews and resumes.
Why it matters here: the market already believes in the value. The gap is a reliable way to build these skills in the team you have — you can’t consistently hire what you can’t assess.
2 in 3
of all jobs will be in soft-skill-intensive occupations by 2030
Deloitte Access Economics — Soft Skills for Business Success
Deloitte forecasts soft-skill-intensive occupations will account for two-thirds of all jobs by 2030 (up from half in 2000), growing at 2.5× the rate of other occupations — driven by digital disruption, globalization, and demographic shifts.
Why it matters here: the demand curve only steepens. Building these skills in your team — or in yourself — is getting ahead of where every job is going.
70M
job transitions analyzed: foundational skills beat specialized expertise in the AI era
Hosseinioun, Neffke, Youn & Zhang — Harvard Business Review (2025)
A computational study of 1,000+ occupations and 70 million job transitions found that foundational abilities — collaboration, adaptability, structured thinking — matter more for individuals and companies than specialized expertise as AI reshapes work.
Why it matters here: AI raises the premium on the human layer rather than lowering it. The skills that survive technological churn are the ones the method trains.
95%
of enterprise generative-AI pilots deliver no measurable return
MIT NANDA — The GenAI Divide: State of AI in Business (2025)
MIT’s 2025 study found the overwhelming majority of enterprise GenAI pilots produce no measurable P&L impact. AI makes teams faster — including faster at the wrong work.
Why it matters here: tools amplify judgment, good or bad. The vision, values, and clarified definition of victory have to come first — that’s the layer the People First Method builds.
2×
AI projects fail at twice the rate of non-AI technology projects
RAND Corporation (2024) · S&P Global (2025) · Gartner (2025)
RAND found more than 80% of AI projects fail — double the rate of conventional IT — for the same root causes: misunderstood problems, unclear business value, and chasing the technology over the use case. S&P Global found companies abandoning most of their AI initiatives jumped from 17% to 42% in a single year; Gartner predicts over 40% of agentic-AI projects will be canceled by end of 2027.
Why it matters here: AI doesn’t just fail to fix why projects fail — it raises the stakes. The causes are the same people, clarity, and governance gaps the People First Method trains against.
19%
slower — experienced developers using AI tools, while believing they were ~20% faster
METR — Randomized Controlled Trial (2025)
In a randomized trial of 246 real tasks, experienced developers took 19% longer when allowed to use AI tools — yet forecast a 24% speed-up beforehand and, even afterward, estimated they had been 20% faster. A measured gap between feeling productive and being productive.
Why it matters here: AI makes work feel faster — including the wrong work. Drucker’s warning, quantified: efficiency is no substitute for judgment about what should be done at all.
~95%
failure rate on the largest, most complex technology projects
Standish Group — CHAOS Report (2020)
66% of technology projects fail overall, and the rate climbs with size and complexity — from ~45% to as high as ~95% on the largest initiatives. Post-mortems point overwhelmingly to non-technical causes: people, communication, requirements, governance, and change.
Why it matters here: this is the problem itself. The failures are people failures — which is why a people-first method moves the number.
66 days
median time for a new behavior to become automatic — and spaced learning beats cramming
Lally et al. (2010) · Cepeda et al. (2006)
Lally and colleagues found new habits take a median of 66 days to become automatic. Cepeda’s meta-analysis of 254 studies found learning spaced over time reliably beats massed practice for long-term retention.
Why it matters here: this is why PFM runs as a months-long cohort with live calls — you can’t binge mindset and behavior change into a team or a career.
What it adds up to
The argument, end to end.
- Technical projects fail at staggering rates — and the causes are overwhelmingly non-technical. (Standish)
- AI is raising the stakes, not lowering them — AI projects fail at twice the rate, abandonment is climbing, and the tools make teams feel faster even when they aren’t. (RAND, S&P Global, Gartner, METR, MIT NANDA)
- The skills that prevent those failures are rising in value across the entire economy — fastest where they combine with technical skill. (Deming, Deloitte, HBR)
- Leaders already believe it — but can’t reliably hire for it, and usually discover the gap after the fact. (LinkedIn)
- The skills are trainable — at any level — with measured ROI. A randomized trial put the firm’s return at 258% in eight months. (Adhvaryu, Kala & Nyshadham)
- Durable change takes spaced practice over months, not a binge course. (Lally, Cepeda)
The People First Method is the system built on that evidence: a months-long cohort that trains the trainable, highest-leverage skills in technical delivery — proven across 30 private cohorts and 1,000+ projects at a ~95% success rate.
Put the evidence to work.
Bring the method to your team, or join the next cohort yourself.
For your team Join a Cohort