Why Big 4 AI Projects Stall Before Production

Why long consulting timelines happen

Long consulting timelines are usually a function of the delivery model rather than the intrinsic difficulty of the workflow.

Consultancies bill across time, staffing layers, and program phases. That commercial structure often rewards duration and overhead more than focused production execution.

This isn't malice. It's incentive structure. When your revenue model rewards duration, you build processes that extend duration. Every Big 4 firm has optimized its methodology for maximum billable hours, not maximum speed.

Where consulting-heavy delivery slows down

Here's what a typical Big 4 AI project timeline looks like for BFSI. We've seen this pattern across multiple clients who came to us after going through (or starting) the consulting route.

Discovery and assessment

A consulting team can spend a long early phase understanding the current state, interviewing stakeholders, mapping processes in slides, and producing large assessment documents before working software appears.

Most of this is the same document they produced for the last banking client, with your company's name swapped in. The processes for KYC onboarding at an Indian brokerage are roughly the same everywhere — CKYC registry, PAN/Aadhaar verification, PEP screening, risk assessment. An experienced regulated-workflow delivery team can often move much faster because discovery is tied directly to system design rather than expanded into a standalone billing phase.

Solution design

Another substantial phase often goes into architecture blueprints, technology diagrams, and specification documents that may later be reinterpreted by a separate implementation team.

The architecture document serves two purposes: it justifies the consultancy's technical expertise, and it creates a contractual baseline for scope management. What it doesn't do is test whether the proposed solution actually works with your data, your systems, and your edge cases.

Proof of concept

Now the A-team builds a POC. This is usually the best work in the entire engagement — senior engineers, focused scope, clean demo. The POC works. Everyone is excited.

Then the A-team moves to the next sales opportunity. The POC enters the graveyard.

Implementation and handoff

A different team — usually more junior — takes over to build the production system. They spend time understanding the POC code and reworking assumptions that were never meant for production. They discover that the POC cut corners that won't work in production. They rebuild significant portions.

This is where compliance becomes a crisis. The POC didn't account for RBI FREE-AI requirements. The data handling doesn't meet SEBI guidelines. The audit logging is missing. Now the team has to retrofit compliance into an architecture that wasn't designed for it.

RBI's FREE-AI framework includes detailed recommendations that affect AI system architecture. If these aren't baked in from day one, you're looking at a significant re-architecture — which is exactly what happens in month 6.

Testing and UAT

Testing expands when the system was not designed for reviewability and production controls from the start. User acceptance testing reveals edge cases that the POC never covered. The compliance team flags issues that need architectural changes.

By the end, you may have a system that technically works but has been through so many handoffs, redesigns, and compromises that confidence in the operating model is weak.

The 4 structural problems

The timeline isn't long because AI is hard. It's long because the consulting model has four structural problems that AI projects make worse.

1. The handoff tax

Every handoff — from sales to discovery, discovery to design, design to POC, POC to implementation, implementation to testing — loses context. Each new team needs time to get up to speed, and repeated handoffs create real delivery drag.

At Aikaara, the same team that scopes the project builds it and deploys it. Zero handoffs. Zero context loss.

2. The document tax

Big 4 projects produce documents: assessments, blueprints, specifications, test plans, status reports. Each document takes weeks to create and weeks to review. Documents are how consultancies demonstrate progress and manage scope.

But documents don't process KYC applications. Working software does. Every week spent writing a specification is a week not spent building the system that generates ROI.

3. The compliance retrofit tax

Consultancies treat compliance as a review gate, not a design principle. They build the system, then run it through compliance review, then fix what compliance flags. This cycle repeats 2-3 times, adding 4-8 weeks.

The alternative: build compliance into the architecture from day one. When we built Centrum Broking's KYC system, RBI/SEBI compliance wasn't a phase — it was the foundation. Every data flow, every decision point, and every audit trail was designed with reviewability in mind before business logic expansion.

4. The generalist tax

Big 4 consultants are generalists. The same team that did your AI project did an ERP migration last quarter and will do a cloud migration next quarter. They're smart people, but they don't have deep BFSI AI expertise.

This means they learn your domain on your dime. They figure out CKYC registry integration for the first time during your project. They discover RBI FREE-AI requirements midway through implementation. Everything that a specialist knows on day one, a generalist discovers in month three.

How focused production delivery actually works

When we talk about faster production delivery, the point is not cutting corners. It is cutting overhead.

The structural contrast

Centrum Broking remains a valid proof point for regulated KYC and onboarding automation, but the safe public lesson is qualitative: governed production delivery matters more than consulting theatre. Read the full case study.

TaxBuddy is the strongest verified proof point on the site: a production AI workflow that achieved 100% payment collection last filing season. Read the full case study.

When Big 4 makes sense (and when it doesn't)

To be fair: Big 4 consultancies have their place. If you're a very large bank with a massive workforce and need enterprise-wide AI transformation across many processes, you might need the organizational change management that consultancies provide.

But if you're a mid-market brokerage, NBFC, or insurance company that needs one or two AI systems in production — KYC automation, document processing, compliance monitoring — the Big 4 model is structural overkill. You do not need a large consulting team and extended program overhead when the real need is a focused governed production system.

The question is not whether a large consultancy can build it. The question is whether you want a consulting-heavy operating model when a tighter governed production path may fit the problem better.

What to ask your current vendor

If you're currently in a Big 4 engagement (or evaluating one), ask these questions:

• "Will the team that scopes this project also build it?" If not, you're paying for the handoff tax.
• "How is RBI FREE-AI compliance addressed in the architecture?" If they don't mention it until the compliance review phase, you're paying for the retrofit tax.
• "When will we see working software processing real transactions?" If the answer is "after UAT in month 7," you're paying for the document tax.
• "What's the fixed price?" If there isn't one, you're paying for the duration tax.