The Short Answer
BCG reports that applied AI is 10 to 15% of the AI-tagged workforce at technology services firms, against 90 to 95% at AI-native firms, and calls this a talent decision rather than a shortage. Buyers should ask who is staffed on their project and what stays behind. With ibl.ai you own all the code and the data when the engagement ends.
Shortening this to "only 10% of 50,000 engineers do applied AI" gets the finding wrong. The real one tells a buyer more, because it separates AI skill on a CV from AI work that reaches production.
What did BCG actually find about applied AI talent at services firms?
BCG found that most AI-tagged people at traditional services firms work on infused AI, meaning AI added to existing work, rather than on applied AI. IANS reporting on the October 29, 2026 report puts infused AI at 85 to 90% of the AI-tagged workforce there.
The analysis drew on more than 2,500 AI-tagged job postings and over 50,000 professional profiles across more than 15 firms.
The report is BCG's The Talent Is Already on Your Payroll. The Clock Isn't. We read its public summary; the detailed figures here come from press coverage.
Note the denominator. The 10 to 15% is a share of the AI-tagged workforce at these firms, not of all engineers and not of the 50,000 profiles. Note the scope too: BCG files it as a local article under an India URL, so check the sample before applying it everywhere.
| Measure | Technology services firms | AI-native firms |
|---|---|---|
| Applied AI share of AI-tagged workforce | 10 to 15% | 90 to 95% |
| Forward-deployed share of applied AI talent | Under 5% | Around 45% |
Is the applied AI bottleneck a talent shortage?
BCG says the bottleneck is not a shortage.
Sreyssha George, a BCG managing director and partner, is quoted saying that "technology services firms do not have an AI talent shortage; they have a talent decision to make," and that "the people who can deliver applied AI are largely already on the payroll."
The report identifies 40 distinct applied AI roles, each linked to an existing technology role. Forward-deployed engineers, semantic engineers and AI operations engineers are among them, as reported.
The moves are not equally easy. Going from data scientist to forward-deployed engineer is reported as roughly three times the effort of going from data engineer to AI data engineer. Some of that reskilling is quick, and the part closest to customers is not.
How big is the AI services opportunity BCG describes?
The opportunity is reported as $100 to 200 billion in net new work by 2030, not a flat $200 billion. AI-led services are expected to add $300 to 400 billion of growth, offset by around $200 billion of existing contract revenue.
So $200 billion is closer to the amount AI-led work takes away than to the prize. The reported services market grows from $1.6 to 1.7 trillion in 2025 to $2.2 to 2.4 trillion by 2030.
That shift explains the incentive problem for buyers. George says clients are moving from paying for hours to paying for outcomes, which changes how work is staffed and rewarded. A firm still paid for hours has a reason to move slowly.
Why should a buyer care about forward-deployed share?
A buyer should care because forward-deployed share is a rough proxy for how much of a firm's AI talent sits next to real systems. Around 45% against under 5% is a large gap, and the forward-deployed model exists because the hard work is local.
Connecting an agent to your identity system, your records and your permissions cannot be done from a slide. It takes someone reading your schema and sitting with your team.
BCG's data does not prove that services firms do this badly, only that few of their applied AI people carry that role.
What should a buyer ask an AI services firm before signing?
A buyer should ask five questions that turn a capability deck into something checkable.
- Name the people on our project. How many hold applied AI roles rather than infused AI roles?
- How many will work inside our systems, and for how long?
- Are we paying for hours or for a defined outcome with acceptance criteria?
- Which deliverables do we keep: source code, infrastructure definitions, evaluation harnesses?
- Who at our organization can run this without you, and when?
The last two matter most, because hiring a services firm moves the talent gap rather than closing it unless skills and code stay behind.
How does ibl.ai work with a buyer's own engineers?
ibl.ai engineers build alongside your team, and the code stays with you. With ibl.ai you own all the code and the data, on your infrastructure, model-agnostic across any LLM, with no per-seat pricing.
Forward-deployed engineering at ibl.ai delivers connector source code, infrastructure manifests, policy configuration and evaluation harnesses, so your team can run and extend the system after handover.
That matters for public buyers too, where owning code is only useful if someone can operate it.
ibl.ai is family-owned and operated from New York, NY.
Source: figures and quotes are from press coverage of BCG's report, including IANS via The Hans India. We read BCG's public summary, not the full report, so check it for exact definitions before citing these numbers.