# AI Consulting in Practice

NLW, AI Daily Brief and Super.ai | AI Engineer CODE 2025 | 18:18

Source: https://www.youtube.com/watch?v=ehQFj6VmuI8
Channel: AI Engineer (https://www.youtube.com/@aiDotEngineer). Summarised by AIE Talks.
Page: https://aietalks.com/talks/ai-consulting-in-practice
Published: 2025-12-18
Tags: enterprise, product-strategy, team-adoption

## TL;DR
- Enterprise AI adoption is growing, but most organizations are still experimenting or running pilots rather than operating AI at scale.
- Time savings is the most common reported benefit, while automation and agent use cases report higher self-reported ROI than basic productivity use cases.
- Organizations that pursue AI across multiple use cases and functions tend to report better ROI than organizations running isolated experiments.

## Summary
NLW combines a macro view from hosting AI Daily Brief with a closer view from interviewing executives through Super.ai. He describes enterprise adoption as widespread, with coding and software engineering seeing especially strong growth. Agents have moved into production at more large companies, yet most organizations remain in pilot or experimentation phases. The main part of the talk presents early results from a self-reported ROI study covering more than 2,500 use cases. Time savings dominates, but respondents also report gains in output, quality, capabilities, decisions, revenue, and risk reduction. NLW is careful about the sample bias, since participants are engaged with an AI podcast and volunteered their answers. Even so, the results show strong optimism, especially around automation and agents. He argues that organizations get more value when they pursue several connected use cases across departments instead of treating AI as a collection of isolated experiments.

## Key ideas
### Enterprise AI adoption is broad, but scaled deployment remains uncommon
[01:46](https://www.youtube.com/watch?v=ehQFj6VmuI8&t=106s)
NLW says almost every enterprise is using AI at least a little, with a major increase in coding and software engineering use. The expected wave of broad automation has not arrived across every function, but agents have still moved into production faster than many people expected. KPMG's survey of companies with more than a billion dollars in revenue found that the share with full-production agents rose from 11% in Q1 to 42% in Q3. At the same time, McKinsey's study found that only 7% of organizations considered themselves fully at scale, while about 62% were still experimenting or piloting.

### Large organizations have often adopted AI faster than smaller ones
[04:19](https://www.youtube.com/watch?v=ehQFj6VmuI8&t=259s)
NLW challenges the assumption that small companies automatically move faster because they are more nimble. Across the past two years, he says the biggest organizations have often made the largest AI adoption efforts. Adoption is uneven across departments, though. IT operations is pulling ahead of other functions. The companies leading overall are doing more than isolated experiments. They are building an organization-wide strategy, running multiple efforts at once, and considering revenue growth, new capabilities, and new product lines alongside productivity and time savings.

### Companies expect to spend more while demanding better evidence of impact
[06:06](https://www.youtube.com/watch?v=ehQFj6VmuI8&t=366s)
The organizations in KPMG's survey planned to increase their AI spending over the next year, from $114 million at the beginning of the year to $130 million in the latest study. Deloitte's data showed that more than 90% intended to increase spending. Expectations for returns have also moved forward. In the prior year's KPMG CEO survey, 63% expected to realize ROI in three to five years. In the newer survey, 67% expected it in one to three years, while 19% expected returns within six months to one year. NLW says traditional impact measures are struggling to describe AI investments.

### The early ROI survey reports positive results, with important sample limits
[08:41](https://www.youtube.com/watch?v=ehQFj6VmuI8&t=521s)
NLW describes a survey that collected more than 1,000 organizational submissions covering roughly 3,500 use cases, with his first analysis focused on about 2,500. Respondents selected a primary benefit from categories such as time savings, increased output, quality, new capabilities, decision-making, cost savings, revenue, and risk reduction. About 44.3% reported modest ROI and 37.6% reported high ROI, while about 5% reported negative ROI. NLW warns that the audience is self-selected, since participants listen to an AI podcast and volunteered their information. Even respondents with negative ROI were generally optimistic about future improvement.

### Time savings is the common starting point for AI use cases
[10:35](https://www.youtube.com/watch?v=ehQFj6VmuI8&t=635s)
Time savings accounts for about 35% of the submitted use cases. Reported savings cluster between one and ten hours, especially around five hours per week. NLW puts that figure in practical terms: recovering five or ten hours each week can return roughly seven to ten work weeks over a year. He says this matters to enterprises even when their larger goal is to create new capabilities. Other common categories include increased output and improved quality, so the reported benefits extend beyond simply reducing the time spent on existing work.

### ROI patterns differ by organization size and seniority
[11:44](https://www.youtube.com/watch?v=ehQFj6VmuI8&t=704s)
Organizations with 200 to 1,000 employees have a larger share of use cases aimed at increasing output. NLW suggests these companies may have reached a meaningful scale while still seeking ways to expand their capabilities, although he has not established why the pattern exists. Executives and senior leaders report less focus on time savings and more focus on increased output and new capabilities. They also report more transformational impact. Seventeen percent of use cases submitted by leaders were described as already having transformational impact and ROI, though NLW notes that leadership roles may naturally select more ambitious use cases.

### Risk reduction is rare as a primary benefit but often highly transformative
[14:01](https://www.youtube.com/watch?v=ehQFj6VmuI8&t=841s)
Only 3.4% of surveyed use cases selected risk reduction as their primary ROI category. Those use cases were the most likely to be described as transformational, with 25% receiving that rating. NLW connects this pattern to compliance, risk, and back-office work, where large volumes of information create persistent operational problems. AI can help with that volume. He also reports that healthcare and manufacturing use cases showed higher average impact than the overall sample, although technology and professional services made up a large share of the submissions.

### Automation, agents, and broader programs report stronger ROI
[15:59](https://www.youtube.com/watch?v=ehQFj6VmuI8&t=959s)
Use cases that mention automation or agents report much higher self-reported ROI than the average. NLW presents this as the next layer beyond common first-tier applications such as producing more content, improving content quality, and recovering work hours. He also finds a direct relationship between the number of use cases an organization submits and the ROI it reports. He gives several possible explanations, but says the pattern supports moving beyond one-off experiments toward coordinated work across functions and disciplines.

## Notable quotes
- "The short of it is enterprises are adopting AI in a growing fashion. Pretty much everyone is using it at least a little, and increasingly they're using it a lot." (01:46)
- "Only 7% of the organizations that they talk to claim or see themselves as fully at scale with AI and agents." (04:22)
- "Time savings is the default. It's the starting point for so many organizations." (10:35)
- "The more use cases that a person or an organization submitted, the better they tended to see ROI for." (16:39)

## Tools & references mentioned
- AI Daily Brief
- Super.ai
- KPMG
- McKinsey
- Deloitte
- MIT
- roicervey.ai

## Who should watch
- You are deciding whether your company's AI work should stay focused on isolated productivity pilots or expand across departments.
- You need early benchmarks for where organizations report ROI from coding, automation, agents, risk work, or time savings.
- You want a candid view of enterprise adoption that separates production agents from pilots and acknowledges self-selection in ROI data.

## Related talks

- [Anthropic in the Enterprise](https://aietalks.com/talks/anthropic-in-the-enterprise) (Alexander Bricken & Joe Bayley, Anthropic, 20:55)
- [Leadership in AI Assisted Engineering](https://aietalks.com/talks/leadership-in-ai-assisted-engineering) (Justin Reock, DX, 18:11)
- [Can You Prove AI ROI in Software Engineering?](https://aietalks.com/talks/can-you-prove-ai-roi-in-software-engineering) (Yegor Denisov-Blanch, Stanford, 16:40)
- [The ROI of AI: Why You Need an Eval Framework](https://aietalks.com/talks/the-roi-of-ai-why-you-need-an-eval-framework) (Beyang Liu, Sourcegraph, 25:28)
- [Moving away from Agile: What's Next](https://aietalks.com/talks/moving-away-from-agile-whats-next) (Martin Harrysson & Natasha Maniar, McKinsey & Company, 21:55)