Three-column diagram showing source engineering roles on the left, bridge roles in the middle, and target GenAI roles on the right, with curved lines connecting valid transition paths.
YOUR CURRENT ROLE
BRIDGE STEP
GENAI ROLE
Selected pathway
Software and backend engineers most commonly transition into GenAI Developer or GenAI Architect roles, via an AI/MLOps Engineer or AI Software Architect bridge. This is the largest single pathway in GCC GenAI hiring today.
Two roles fall outside the map
Prompt Engineer and AI Data Curator have no engineering transition path. They must be sourced from outside the existing IT workforce — making them the hardest roles for GCCs to staff today.
Source: Zinnov analysis, 2025.
The two roles without a transition pathway are the harder problem. Prompt Engineer and AI Data Curator have no engineering origin. They must be sourced from outside the existing IT workforce; from linguistics, product design, technical writing, domain expertise. They are also, not coincidentally, the two roles that GCC hiring leaders consistently report as the hardest to fill. The roles without a transition pathway are the ones the existing recruiting playbook cannot find at all.
Why GCCs Can’t Hire for These the Old Way
Three structural forces make the old hiring playbook unworkable.
No degree pipeline exists anywhere in the world:
There is no “BTech in Prompt Engineering.” There is no “MS in Synthetic Data Generation.” The 8 roles are too new, the underlying tools change too quickly, and the curricula have not had time to catch up.
Even where universities have launched AI programs, the curricula focus on foundational ML and deep learning, not on the production Generative AI stack GCC teams are hiring for. A graduate of even a top Indian engineering programme in 2026 will have spent more time on classical Machine Learning than on transformer architectures, more time on Hadoop than on vector databases, more time on monolithic ML systems than on agent frameworks.
The degree filter, applied to these roles, screens out almost everyone, including the people who can actually do the work.
The skills are arriving from non-traditional paths:
Roughly 70% of new technical learners did not take a traditional CS degree path. Mid-level technical roles in Cybersecurity, Data, DevOps, and QA are shifting downward into the early-career bands, their share of early-career postings rose from approximately 14% in 2022 to 22–28% in 2025.
This is the structural consequence of AI tools amplifying junior productivity.
A second-year engineer with strong tooling can now deliver what a fifth-year engineer delivered three years ago. The hiring filters that assume “years of experience” map cleanly to “capability” no longer hold.
Demand is outpacing supply by an order of magnitude:
Approximately 60% year-on-year demand growth for niche AI roles is colliding with a structurally constrained talent supply.
The market is already pricing the gap. BTech graduates with niche AI/ML skills earn 1.2–1.7x more than peers with mainstream skills. 7% of GCCs are offering retention bonuses up to 30% of base pay for niche skill holders. AI/ML engineering talent globally grew 20% faster than software engineering talent in 2024. And approximately 63% of software development lifecycle roles are on the verge of significant AI automation, with 64% of manual SDLC effort projected to be eliminated by GenAI.
The roles that need to be filled are not the roles the existing talent system is producing.
What Leading GCCs Are Doing Instead
The good news: leading GCCs are already adapting. Four sourcing patterns are visible in the data.
Skills-based hiring pilots
31% of entry-level GCC hires in 2024 were assessed via certifications or skills tests, up from 19% in 2022. Approximately 45% of Indian GCCs plan to move away from degree requirements entirely.
The trajectory is set. The pilots that work share a common pattern: replace the degree screen with a project portfolio review, replace the aptitude test with a hands-on technical assessment using real tools, replace the panel interview with a simulated work scenario. Predictive analytics applied to these competency-based screens delivers approximately 20% improvement in new-hire quality over traditional methods. The data on what works is no longer ambiguous.
Internal transitions and reskilling at scale
The transition pathways are concrete. A Backend Engineer can transition into an AI Software Architect, then into a Generative AI Architect. A QA Engineer can transition into AI-Driven Automation Testing, then into a Control Models Specialist. A Data Scientist can transition into an NLP Engineer or ML Engineer, then into an AI Bias Expert or Synthetic Data Engineer.
The existing GCC workforce is the most underused source of GenAI talent in India today. Building the internal transition pipeline is faster, cheaper, and lower-risk than competing with every other GCC for the same scarce external candidates.
Acquisitions as a recruitment channel
Adobe acquired Rephrase.ai. IBM acquired Prescinto. Honeywell acquired Flutura. 3 Bengaluru AI start-ups, all acquired in 2024–25, all primarily for the talent. M&A has become the de facto recruitment strategy for frontier teams that cannot be built through traditional hiring on any reasonable timeline.
The Hiring Playbook GCCs Need Now
Three concrete shifts will determine which GCCs successfully staff for the next wave of AI work.
Replace the degree filter with a skills filter
The old hiring funnel has three stages, and each of them is screening out candidates who could do the work. The skills-based funnel replaces each stage with something that measures capability instead of credential.
Stage 1
Degree screen
BTech / MTech required
What it filters for
Engineering degree from a recognised institution, minimum CGPA cutoff.
What it misses
The 70% of technical learners without a traditional CS degree path. Prompt engineers from linguistics. AI persona designers from creative writing. AI bias experts from social sciences and law. None of these candidates make it past the first screen.
Stage 2
Aptitude test
Generic reasoning / quant
What it filters for
General problem-solving ability tested through abstract questions: logical reasoning, quantitative aptitude, pattern recognition.
What it misses
Practical capability with the actual tools the role requires. A candidate who scores 99th percentile on abstract reasoning may have never touched PyTorch, LangChain, or a vector database. The test rewards test-takers, not practitioners.
Stage 3
Panel interview
Structured Q&A
What it filters for
Verbal articulation, confidence under pressure, ability to produce answers to structured questions within a fixed time. Often evaluated on whether the candidate reaches a specific “right” answer.
What it misses
How the candidate actually thinks through unfamiliar problems. Articulate candidates with weak technical judgement pass. Strong technical candidates who need time to reason can fail. The interview measures interview skill, not job skill.
Stage 1
Portfolio review
GitHub, model cards, prototypes
What it filters for
Evidence of shipped work. GitHub repositories showing commits to real projects. Hugging Face model cards. Deployed prototypes. Technical blog posts.
What it gains
Candidates are evaluated on what they have built, not on where they studied. Non-traditional learners with genuine skill become visible for the first time. The signal is the artifact, not the credential.
Stage 2
Hands-on assessment
Real tools, real problems
What it filters for
Ability to use the actual production stack: PyTorch, LangChain, vector databases, foundation model APIs. Candidate is given a realistic technical problem to solve in a working environment.
What it gains
Assesses the skill the job actually requires. Eliminates the gap between “can pass the test” and “can do the work.” Practitioners outperform test-takers, which is what you want.
Stage 3
Simulated task interview
Evaluated on approach
What it filters for
The candidate is given a realistic problem and a working environment, and evaluated on their approach: how they diagnose, what tradeoffs they consider, how they recover from dead-ends, how they communicate through the process.
What it gains
Approximately 20% improvement in new-hire quality over traditional methods. Measures actual problem-solving, not interview performance. Favours candidates who work like the job actually works.
Source: Zinnov analysis, 2025.
None of these changes require new technology. All of them require the hiring team to give up the false comfort of the degree credential.
The transition is already underway. 31% of entry-level GCC hires in 2024 were assessed via certifications or skills tests, up from 19% in 2022. Approximately 45% of Indian GCCs plan to move away from degree requirements entirely. The data on what works is no longer ambiguous.
Build the internal transition pipeline before you need it
Map every adjacent role in your existing engineering organisation against the eight new AI jobs.
- Software Engineers and Backend Engineers map most naturally to Generative AI Developer and Generative AI Architect.
- Data Scientists and ML Engineers map to AI Bias Expert and Synthetic Data Engineer.
- QA Engineers map to Control Models Specialist.
- UX Designers and content strategists map to AI Persona Designer.
Build the transition curriculum now, while the urgency is moderate. 6 months from now, when the urgency is acute, the curriculum will already be operational and the first cohort will already be productive.
The GCCs that wait for the talent to appear externally will be the ones still looking 18 months from now.
Co-fund external upskilling for non-employees.
- Accept stackable micro-credentials as proof of capability in your hiring funnel.
- Integrate FutureSkills Prime credentials into your applicant tracking system as first-class signals.
- Partner with state skill missions to seed the talent pool you will hire from in 12 to 24 months.
The L&D budget; ₹17,000 to ₹26,000 per employee per year on average across Indian GCCs, is currently spent almost entirely on existing employees. Reallocating even 15-20% of that spend toward pre-employment skilling expands the addressable talent pool meaningfully and creates a credentialed pipeline that competitors cannot easily replicate.
The Window Is Narrow
The forecast that India will add over 500 new GCCs by 2030 assumes the talent system can absorb them. That assumption is contingent on the GCC hiring playbook adapting to the 8 new AI jobs; and to the dozens of additional roles that will follow them as the GenAI stack continues to evolve.
The window for adaptation is narrow. The MNCs evaluating right-shoring decisions in 2026 and 2027 will weigh which locations can credibly staff frontier AI teams at scale.
The 8 new AI jobs are the test case. Get the hiring model right for these roles, and the rest of the talent strategy follows. Get it wrong, and the lead will erode role by role, charter by charter, until the next wave of AI work goes somewhere else.
The framework, the data, and the policy infrastructure are all in place. What remains is the operational decision, at every GCC, by every hiring leader, to actually change the playbook.
Zinnov proprietary data. The data used in this blog has been referenced from multiple Zinnov and Draup reports, forums, surveys, and conversations with GCC Heads.