How will GenAI impact IIQ?
5 to 20 minute reading time. Click on each question to see the key facts, modify the charts to capture your opinion.
How are BFSI analytical jobs exposed to AI disruption?Financial analysis is one of the most AI-exposed occupations in the economy, and every serious measure agrees despite using different methods. Two axes frame it: across occupations, financial analysis sits in the top exposure tier; within it, the procedural execution layer is far more exposed than the research and judgement layer. Exposure here means the share of tasks AI can do, not jobs already lost; the studies that look further find the most-exposed workers are also among the best placed to adapt.
Key facts · 3
Every serious measure puts financial analysis in the top exposure tier, by a different route
| Study (method) | What it measures | Where financial analysis lands |
|---|---|---|
| Anthropic Economic Index, 2026 | Real AI usage mapped to tasks (actual coverage) | Top-10 occupation, ~57% task coverage (programmers ~75% lead) |
| Goldman Sachs, 2026 | Automatable-task model + firm adoption | Finance a higher-exposure sector; ~2.5% near-term automation risk economy-wide |
| NBER, Manning & Aguirre, 2026 | Exposure ranking across occupations (95.9% of the workforce) | In the top exposure quartile (exposure >= 46%) |
| Older indices: Eloundou 2023, Felten / AIOE 2021-23, OECD 2023 | Theoretical task and ability exposure scoring | Consistent: financial analysts near the top of every ranking |
Several studies, built four different ways, reach the same place. The newest read real behaviour or model the economics: the Anthropic Economic Index measures what AI is actually used for from platform data, and Goldman models the automatable share of tasks and how fast firms adopt. NBER ranks the same occupations by exposure. The older indices score what AI could do, task by task. They disagree on magnitude and even on what 'exposed' means, yet financial-analysis work lands in the top tier on all of them. The IMF puts the economy-wide baseline at about 60% of advanced-economy jobs exposed; financial analysis sits well above it.
These are potential-and-usage measures, not headcount. None shows a rise in unemployment for high-exposure roles through 2025-26, and the OECD notes the most-exposed occupations are not the same as the most-automatable. What happens to jobs is the 'Net job loss or gain?' question.
But it is the procedural layer that is exposed, not the judgement layer
India offshore work by type (2026)
What decides exposure is the type of work, not the job title. Zinnov sorted more than 1.7 million job descriptions from India's Global Capability Centres (the offshore units that banks and corporates staff and run themselves, also called captives) into four kinds of work. Two are exposed: routine standardised tasks, and the skilled but repeatable execution of known processes. Two are protected: work that turns on deep experience and judgement, and frontier research. These figures cover all offshore knowledge work, not financial analysis on its own, but the same dividing line runs through it: spreading and monitoring are exposed, original judgement is not.
The mix is shifting towards safer work: routine tasks have roughly halved since 2015 and deep-judgement work has nearly doubled, to 38%. Even so, the exposed half is still the larger one: 54.9% of the work in India against 42.9% at the onshore head-office sites.
Mapped onto IIQ's book, exposure runs above the GCC average
| IIQ practice | Dominant work | Exposure |
|---|---|---|
| Operations & support | Document processing, customer checks (KYC) | Highest |
| Fundamental research | Earnings, comps, pitchbooks | High |
| Credit & lending | Origination, monitoring | Moderate-high |
| Quantitative research | Model dev and validation | Lower |
IIQ's AI talent strategy maps the exposure of each practice. IIQ is research-led relative to the typical GCC commodity mix, so its exposure sits above the GCC mean on axis one. On axis two, the same split runs through every IIQ practice: spreading, extraction and routine monitoring are the exposed layer, while credit judgement, model design and thematic research are the protected layer.
The protected slice is real and defensible, which is exactly where the strategy should concentrate IIQ's people.
Share of BFSI analytical tasks automatable with current GenAI tools
Starting bars reflect the per-process effort-reduction evidence (equity earnings and comps ~75-85%, credit monitoring ~80%, model validation ~65%), weighted down by each role's judgement share. Adjust to your own view.
How fast will the transition happen?The technology is ready ahead of adoption: capability moves at the speed of research, deployment at the speed of risk committees. That gap is the strategic uncertainty.
Key facts · 4
Adoption is already broad, but shallow
Wide use, shallow production deployment
About 49% of US jobs have used AI for at least a quarter of their tasks, yet only about 31% of enterprises run any AI agent in production. Use is wide; deep production deployment is not.
Shallow adoption can deepen fast. Gartner expects about 40% of enterprise apps to feature task-specific AI agents by 2026, up from under 5% in 2025, so the floor rises without a committee deciding anything.
In banks specifically, use is high but agents are barely in production
Banks: high AI use, early agent deployment
The pattern is sharper inside banking. NVIDIA's 2026 financial-services survey finds 65% of firms now actively using AI, up from 45% a year earlier, yet only 21% have an AI agent deployed. Across the 50 largest banks, Evident counts just nine with agents in pilot or production, and only 15 of 97 deployed generative-AI tools face the customer; the rest are internal. Technical readiness is running ahead of production adoption.
This can jump quickly. Evident found new bank use cases doubled in six months and agentic technologists grew more than tenfold; among AI practitioners LangChain already puts most agents in production. Cheap open-source and on-device models, important to more than 80% of firms, let teams ship without a central gate.
Governance is the speed limit
Agent ambition vs governance maturity (2026)
This is exactly the class of risk governance is supposed to catch, and exactly the class that runs through credit, risk, compliance and trading.Xu, Li & Jiang (2025)
What gates deployment is governance, not capability. Deloitte found that nearly three-quarters of firms plan to deploy AI agents within two years, but only 21% have a mature way to govern them. And the failures governance has to catch are subtle: when seven AI tools rewrote 2,245 resumes and then scored them, each heavily preferred its own output even where human judges rated the originals clearer, a fair scorer would land at 50% and none did.
Two things cut the other way: a desktop agent a senior analyst runs without IT, or a capable local model, can bypass the committee and collapse the timeline into a cliff; and the validation work itself is billable activation work IIQ could own, not just a drag on the pace.
Ambition is running well ahead of activation
Ambition vs activation in the enterprise (2026)
Deloitte's 2026 survey of 3,235 leaders across 24 countries, financial services among them, puts numbers on the gap. Sanctioned AI access widened from under 40% to about 60% of staff in a year, yet only 25% have moved 40% or more of their pilots into production and just 30% have redesigned any core process around AI. The ambition is real; the operating capability to deliver it is not there yet.
The pace could surprise in either direction. It could run even slower if deployments keep stalling in pilots; or it could break faster, since capability compounds for early movers (Anthropic) and 54% of firms expect to clear the pilot-to-production hurdle within six months. Committee-paced is the base case, not a certainty.
Share of BFSI analytical jobs materially changed by GenAI over the next five years
The starting bars are anchored to the evidence: year 1 is low because production deployment is still early (about 31% of enterprises run any agent, far fewer have it changing whole roles), rising to ~55% by year 5, the procedural layer AI targets first (Zinnov). Adjust to your own view.
Will AI be used to raise throughput or cut costs?When AI raises productivity, the gain lands as more output, or as fewer people at the same output. The split is a strategic choice.
Key facts · 4
The gain is real and skewed to juniors
Throughput lift concentrated in the least-experienced
Additional resolutions per hour, by pre-deployment skill quintile
Across 5,172 support agents, AI lifted output about 14% per hour. The lift is concentrated in the lowest-tenure quintile and near zero for the highest, so AI compressed the skill gap rather than widening it.
A gain that mostly helps juniors eats into the cost advantage of offshoring (the savings from doing the work in a lower-cost location), because the client's own junior staff capture it too.
Where output is fixed, efficiency becomes cuts
Compliance and regulatory reporting are bounded by mandate. Efficiency there converts directly into savings, not into more work.
Where output is elastic, such as coverage breadth or idea generation, the same efficiency converts into more work.
Clients decide where the gain lands
It's not cost-cutting. It's replacing, in some cases, lower-value human capital with the financial capital and the investment capital we're putting in.Standard Chartered CEO, May 2026
The split is a strategic choice made by the buyer, not a property of the technology.
Other clients explicitly reinvest the dividend into broader and deeper coverage.
IIQ has shown it can capture the gain
IIQ has already demonstrated roughly 30% efficiency on credit-report generation with its own tooling, so the productivity discipline is in its hands.
When the client is billed per analyst (per full-time-equivalent, or FTE), a captured gain leaks straight to the client unless what IIQ sells changes.
Throughput vs cost cutting: the net effect on BFSI analytical workflows
BFSI analytical workflows: Net job loss or gain?Past automation produced both outcomes. The question is which side of the elastic/inelastic line IIQ's book sits on.
Key facts · 5
Two centuries say elastic demand grows jobs
Jobs after the spreadsheet arrived
Approximate change in US employment, data-handling occupations
Burn-Murdoch (FT, 2024); US CPS; framework after Bessen
US data on textiles, steel and autos shows automation often spurred employment because product demand was elastic to lower cost. The spreadsheet is the cleanest finance example: it cut bookkeeping-clerk employment, but grew US financial analysts roughly tenfold and more than tripled accountants, because cheaper routine numbers expanded downstream demand for analysis.
The 1980s computer wave is the honest analogue: bookkeeping clerks, secretaries, operators and travel agents contracted, not expanded. The pattern holds only where demand is elastic, and most of IIQ's book is not.
The computer wave is the honest analogue
From the 1980s, automation produced net contraction in bookkeeping clerks, secretarial roles, telephone operators and travel agents. ATMs did not kill tellers; mobile banking did, two decades later.
AI improves at the very tasks displaced workers would normally redeploy into, which weakens the usual cushion.
The split is already in the labour-market data
US employment, one-year change to May 2025
In the year to May 2025, AI-exposed roles fell while AI-resilient roles rose, and the split runs through the middle of every sector rather than across sectors.
The effect is task-level, not sector-level, so resilient niches survive inside shrinking sectors and vice versa.
US finance is IIQ's cleanest leading indicator
In the same year, US finance clerical roles fell 4.7% while analytical roles grew 2.2%. IIQ serves the analytical layer of those same banks.
Analytical resilience is relative; the floor keeps rising as AI climbs the task ladder.
The dominant book pattern is contraction
For fixed-quantity work, productivity gains become fewer FTEs at the same output. IIQ cannot assume elastic demand will absorb the dividend.
The elastic slice (ideas, deals, themes) is genuine and is where any net growth would come from. And exposed is not displaced: NBER puts financial analysts in the top exposure quartile yet near the 99th percentile on adaptive capacity, against about 22 for office clerks, so the same workers are unusually able to redeploy into the new analytical roles.
Net effect on employment in BFSI analytical workflows, 5-year horizon
BFSI analytical workflows are the roles that produce or quality-check analytical output for banks, asset managers, insurers and corporates: equity and credit research, financial spreading and modelling, regulatory and risk reporting (the Basel and FRTB capital rules, model validation), customer due-diligence (KYC) and data clean-up, and ESG and portfolio analytics. They exclude pure IT, sales and front-office relationship roles. Set the share of today's roles replaced and the share of genuinely new analytical roles created over five years; the net is the difference.
How will offshore-outsourced BFSI analytical jobs be impacted?BFSI knowledge-process outsourcing (KPO) and India's in-house offshore centres are both growing fast. The third-party outsourcing slice IIQ sits in is the squeezed middle.
Key facts · 5
Three markets, three very different trajectories
Annual growth by market
Zinnov-NASSCOM India GCC Landscape FY26; company filings
BFSI KPO is growing at roughly 19% a year and India's in-house offshore centres at 24%. The third-party IT-services slice IIQ competes in grew 0.06%.
The two growing markets are precisely the ones IIQ does not sit in today.
In-house offshore centres are absorbing the resilient work
Headcount at India's in-house offshore centres grew from 1.9 million to 2.36 million in a year, up 24%, against +0.06% at the large services firms. More than 500 of these centres are dedicated to AI and machine-learning work.
Same talent pool, different employer. The work migrated rather than vanished, which leaves a door open for whoever supplies that capability.
BFSI GCC revenue nearly doubled in four years
BFSI GCC revenue jumped from $13.4B to $24.6B, and a meaningful share came at third-party expense.
Part of that is net-new AI demand, not pure substitution of existing third-party work.
The BFSI layoff wave names the function
| Bank | Disclosed cut | Horizon |
|---|---|---|
| Citigroup | 20,000 | by end-2026 |
| HSBC | up to 20,000 | 3-5 years |
| Standard Chartered | 7,800 | by 2030 |
| ANZ | 3,500 | multi-year |
Disclosed cuts run past 85,000 across roughly 15 tracked firms, with offshore hubs most exposed and AI increasingly named as the driver.
Cuts cluster in back office, ops and call centres; research-grade analytical work is less directly named so far.
IIQ's current model is on borrowed time
Unless IIQ starts selling expertise and outcomes rather than headcount, its third-party offshore model stays caught between work moving back onshore (reshoring) and work moving into clients' own in-house offshore centres.
Capability density is exactly where IIQ's domain depth could let it win the resilient slice.
How offshore-outsourced BFSI analytical jobs compare to the overall trend
Which business model?Keep selling analyst time (the full-time-equivalent, or FTE, model), shift to consulting, or become a product company. Selling time by the head is the thing under attack.
Key facts · 4
Business as usual: keep what you sell, change how the work gets done
Keep charging for analyst time and make the analysts AI-ready, capturing the productivity gain inside the existing book of work.
Selling analyst time is squeezed from above by AI replacement and from below by work moving back onshore and into clients' own offshore centres.
Consulting: sell decisions, not hours
Shift towards output-based advisory on AI integration, where the deliverable is a decision rather than a headcount.
IIQ's costs sit almost entirely in billable delivery hours rather than reusable products, roughly 99 to 1. An advisory model does not fit that structure without breaking the keep-staff-billable engine the business runs on.
Product: licence-based
Build and licence tools. Covenant Guard, GenEye Credit and Coverage IQ already point this way.
Becoming a product company is a different business with different economics, incentives and people.
The segmented answer
| Book | Operating model | Why |
|---|---|---|
| Sell-side & buy-side research | Human-led domain expertise | Human-led; the method travels with the people |
| Credit & lending | Platform plus domain expertise | The named platforms are the durable edge |
The brief's own conclusion is a segmented model, not a single pick.
Segmentation needs group-level orchestration that has not been visible to date.
Where should IIQ land?
How will roles transform?Three post-AI roles emerge across the pyramid. The question is whether the reshape is durable or a way-station.
Key facts · 3
The three post-AI roles
| Role | What the human does | Why outsource it |
|---|---|---|
| Feedback provider | Ground truth, accuracy checks, low-confidence validation | Cost saving on expert hours |
| System supervisor | Maintain, fine-tune, handle edge cases, own outputs | Cost saving plus AI-systems know-how |
| System designer | Design the methodology and the automation | Bottleneck: scarce staff |
As automation deepens, the human work moves up the chain through three archetypes.
Each archetype is one the client increasingly prefers to keep close.
Junior to human feedback provider
Displaced juniors move to providing ground truth and quality control on AI output.
Feedback need shrinks as models improve, so the role is a way-station rather than a destination.
Senior to supervisor or designer
Seniors move from running workflows to supervising and designing the automation that runs them.
Supervision and design carry accountability, which is exactly why clients onshore them or pull them into captives within two to five years.
The reshape into feedback and supervision roles is durable, not transitional.
What is IIQ's competitive advantage?Five competitive-advantage dimensions. Only one is a moat IIQ clearly holds today.
Key facts · 3
The five-dimension scorecard
| Dimension | Real | Hollow |
|---|---|---|
| Workflow expertise | The moat; hard to mobilise at scale | Needs AI fluency it lacks |
| Data & ground truth | Useful at design/QC | Mostly the client's |
| Delivery environment | Where gains get captured | The client owns it + the IP |
| Low-cost labour | Real for AI-weak tasks | Cannot win on price |
| Upskilling speed | Possible with a trainable base | Budget/infosec dependent, unproven |
Each claimed advantage has a real version and a hollow one.
Only workflow expertise is a moat IIQ clearly holds, and even that is a moat only once paired with AI fluency.
Hands-on workflow expertise is the moat
The hardest GCC hire in 2026 is the platform engineer who understands credit risk. IIQ's mid-senior band is already domain-literate; the gap is AI fluency, a reskill problem not a hire-from-scratch one.
A moat only if the reskill actually happens against a utilisation-driven P&L.
Data and environment mostly belong to the client
The defensible data sits in the parent group behind separation rules, and inside the outsourcing arrangement itself the data and the delivery environment are the client's.
Access can be improved at the design and quality-control stage, and IIQ's reusable tools, installed inside the client's secure virtual desktop (VDI), capture some of that environment.
How strong is IIQ's edge in AI-enabled delivery today?
Can we grow if the model stays the same?Will new task volume flow to IIQ by default, or will clients keep the gain?
Key facts · 3
The optimistic mechanism is the rebound effect
Broader and deeper workflows, enabled by AI, create net-new demand that flows to the analysts who run the adjacent work.
Rebound requires clients to reinvest the gain offshore, which is a choice, not a default.
The pessimistic mechanism is cost capture
Clients bank the efficiency and reinvest it elsewhere. The work may still exist at market level, displaced out of IIQ's scope.
Market-size forecasts align with the optimistic path, but they describe the market, not IIQ's slice.
In both scenarios IIQ is passive
The outcome turns on client decisions and their drivers rather than on IIQ's own choices.
The active alternative is to change what IIQ sells before client pricing forces the move.
Under business-as-usual, IIQ headcount grows over the next five years.
Same model, optimistic scenarioA rebound: AI expands the book, juniors move to feedback and QC, seniors design systems, net headcount rises.
Key facts · 3
Indexed headcount 100 to about 120 over five years
Optimistic: indexed IIQ headcount
In the optimistic path, demand expansion lifts net headcount even as the task mix changes.
This holds only if demand is elastic and clients reinvest the gain offshore.
Juniors redeploy to feedback and QC
Displaced juniors are reassigned to human feedback and quality control on AI output.
That work can be hired at captives just as easily.
Seniors move to AI system design
Seniors who used to set up workflows now design and build the custom automation.
Design is the role clients least want to outsource.
How likely is the rebound for IIQ specifically?
Same model, pessimistic scenarioClients pocket the gain. Standardised tasks collapse, work migrates to captives and AI vendors, IIQ's scope shrinks about 50%.
Key facts · 3
Indexed headcount 100 to about 50 over five years
Pessimistic: indexed IIQ headcount
In the pessimistic path, standardised work collapses and new-request work declines within IIQ's scope.
The floor depends on how fast IIQ changes the unit of sale, so it is not fixed at 50.
Standardised tasks collapse first
Spreading, extraction and routine monitoring are automated out of IIQ's scope first.
Those are exactly the tasks IIQ can re-own as AI supervision and QC.
The work migrates, it does not vanish
Feedback and AI-design roles get hired at captives and AI vendors instead of IIQ.
IIQ can still compete for the high-expertise fraction of that migrating work.
Clients pocketing the gains is the base case, not the tail.
TrainingExposure is not fluency. Build a structured ladder anchored on workflow apprenticeship, not generic AI literacy.
Key facts · 4
A four-rung persona ladder
| Rung | What they can do |
|---|---|
| User | Use AI tools in their own workflow |
| Power User | Push tools hard, build prompts and checks |
| Builder | Build reusable workflows and skills |
| Enabler | Design systems and train others |
Anchor training on workflow apprenticeship rather than generic AI literacy, across four rungs.
Ladders are easy to draw and hard to fund against a metric that rewards billable hours and treats training time as a cost.
Practice compounds, so start early
The tenure premium in AI success survives full controls. Time on the tools is the asset, which makes early internal exposure a durable edge.
Clients training their own staff erodes the edge, and the window closes as they build the curve in-house.
Where AI actually cuts the hours
Effort reduction once a workflow is redesigned
Redesigned workflows show large, uneven effort reductions across IIQ's own processes.
Under per-FTE billing, those saved hours leak straight to the client rather than to IIQ's margin.
Three pushbacks, each a design principle
The objections are real: clients will do it anyway, it is too expensive, and there is too much uncertainty on what to train on.
Each is real and none is a reason to skip; each instead becomes a constraint that shapes the training design.
How much should IIQ invest in training?
AcceleratorsAI tools, libraries and reusable assets IIQ should build or buy. (Section to be developed.)
Content to be developed.
What should IIQ build or buy first, and why?
AI budget per personHow much per analyst per year on AI tooling and training, and how to measure ROI. (Section to be developed.)
Content to be developed.
What annual AI budget per analyst is justified?