BCG's Build for the Future 2025 study tracked 1,250 companies across 25 sectors and found that only 5% — what BCG calls "future-built" organisations — are generating substantial, compounding value from AI. The study identifies five strategies that distinguish these companies from the 95% that are either generating no value or struggling to scale.
I want to translate each of those five strategies into the specific context of contact-centre and CX AI transformation. Because the abstract playbook — "pursue a multiyear strategic ambition," "reshape and invent with impact" — is insufficient on its own. What matters is what that looks like on a Monday morning in a contact centre in Mumbai or Bengaluru or Gurugram.
Source: BCG Build for the Future 2025, n = 1,250 senior executives.
Move 1: C-suite ownership, not middle-management delegation
BCG's research finds that nearly 100% of future-built organisations have deeply engaged C-level executives. Only 8% of laggards do. Future-built companies are 3x more likely to have appointed a chief AI officer, and 2x more likely to have a chief data officer.
In the contact-centre context, the failure mode here is very specific. AI programmes in contact centres are routinely delegated to people who have the capability but not the mandate.
The head of digital transformation runs a pilot. The CTO approves vendor engagements. The operations director participates in steering committee meetings. And nothing moves at meaningful speed, because no one in the room has the authority to redesign the queue architecture, reallocate headcount, change how supervisors are measured, or commit the capital required to retrain agents at scale.
What C-suite ownership looks like in practice: the COO has a contact-centre AI target on their scorecard — containment rate, cost-per-contact, CSAT — and it is reviewed quarterly in the same meeting where they review the P&L. Not the technology roadmap meeting. The business results meeting.
When I see that, I know the programme will move. When I see AI on the agenda of the "digital transformation committee" — a meeting that the COO attends for fifteen minutes before handing it back — I know it will not.
Move 2: Reshape end-to-end workflows, not isolated automations
BCG's most consistent finding across both the 2024 and 2025 reports: big value comes not from AI pilots or isolated use cases, but from reshaping and reinventing core business workflows end-to-end. Future-built companies are 5x more likely to have AI workflows fully deployed or scaled.
The contact-centre version of this failure is what I call the call deflection trap.
An organisation deploys a chatbot or virtual assistant on their website or IVR. The bot handles some queries. Containment is measured — let's say 30%. Leadership is pleased. The programme is declared a success.
What has not changed: everything downstream. The interactions that escalate to agents are handled the same way they always were. The agent desktop has not changed. The CRM workflow is identical. Supervisor coaching is unchanged. Quality measurement is unchanged. The routing logic is unchanged.
The contact centre has a chatbot, not an AI-transformed contact centre.
A future-built organisation asks a different question: what does the contact centre look like when AI is embedded end-to-end? That means starting with customer intent — before the customer contacts the company — and following the interaction through to resolution. Every step that AI can handle more effectively than a human, handle it. Every step that genuinely requires human judgement, design for human capability.
In practice, this means: the IVR or chat entry point uses conversational AI for intent recognition. Routing is AI-driven, based on predicted issue complexity and agent skill, not static rules. The agent desktop surfaces AI-generated context before the agent speaks. Post-interaction summarisation is automated. Quality monitoring uses AI speech analytics. Coaching recommendations are AI-assisted.
None of those individual technologies is novel. What is rare is deploying them as an integrated system rather than discrete tools purchased from different vendors and integrated superficially.
Move 3: Shared business-IT ownership — and why the IT-led pattern stalls
BCG finds future-built companies are 1.5x more likely to have joint business-IT ownership of AI implementation, with clear decision rights. Exclusive IT ownership is specifically identified as a predictor of stagnating organisations.
This is the most common structural failure I see in Indian enterprise contact-centre programmes. And it has a consistent shape.
IT owns the CCaaS platform contract. IT manages the vendor relationship. IT runs the technical evaluation and selects the product. Operations is "consulted" — they sit in demonstrations, they sign off on requirements documents, they participate in UAT.
But when it comes to the decisions that actually determine whether the programme delivers value — which workflows to prioritise, what success looks like, how agent roles will change, how supervisors will be measured differently — those decisions are not being made. Or they are being made by IT in isolation from the people who understand how the contact centre actually works.
The result is technology that is technically functional but operationally disconnected. An agent assist tool that surfaces recommendations agents have been trained to ignore. An AI routing system that IT maintains but operations does not trust. A quality analytics platform that generates reports nobody reads.
Shared ownership means: the operations director and the IT director both have skin in the game. Operations owns the business outcomes. IT owns the architecture. Neither can make significant decisions without the other. And both are accountable to the same executive sponsor.
Move 4: Workforce transformation first — the 10-20-70 rule in CX
BCG advocates a "10-20-70 rule" for AI transformation: 10% of focus on algorithms, 20% on technology, 70% on people, organisation, and process. Future-built companies invest in broad-based employee enablement — more than 50% of the internal workforce is expected to be upskilled in AI in 2025, versus 20% at laggards.
In a contact centre, this has a specific meaning that is frequently underestimated.
The agent role changes materially in an AI-transformed contact centre. The agent is no longer the primary knowledge retrieval mechanism — the AI handles that. The agent is increasingly the person who handles situations the AI cannot: emotionally complex interactions, edge cases, customers who need human empathy rather than efficient resolution.
That is a different skill set from what most agents were hired and trained for. It requires better listening skills, better emotional intelligence, better judgement about when to de-escalate and when to escalate further. It requires agents who are comfortable working alongside AI suggestions rather than treating them as either gospel or noise.
Most contact-centre AI programmes plan extensively for the technology and almost not at all for this transition. They assume agents will adapt. Some will. Many will not — not because of resistance, but because the training and support to make that adaptation have not been provided.
The organisations that get this right invest in re-skilling before deployment, not as an afterthought. They co-design the new agent experience with agents, not for them. They build supervisory capability to coach AI-assisted agents, which is a different discipline from coaching traditional agents.
BCG's data is clear: companies that involve employees in co-designing AI solutions are twice as likely to achieve smooth adoption. That is not a soft benefit. Adoption is the variable that determines whether the technology investment delivers ROI.
Move 5: Fit-for-purpose architecture — the right stack shape for CCaaS
BCG describes a "hybrid portfolio approach" to agentic AI: a mix of standalone solutions, embedded capabilities, builder platforms, and custom-built components. Future-built companies are 3x more likely to operate a central, integrated AI platform. Only 11% rely primarily on in-house development. Only 4% depend on a single end-to-end vendor.
For a contact centre, the practical translation of this is what I think of as the three-layer stack:
Layer one: The CCaaS platform. This is Genesys Cloud, Amazon Connect, NICE CXone, or one of their significant competitors. The CCaaS platform owns the interaction channel — voice, digital, messaging — the routing logic, and the agent desktop. Choosing the right CCaaS platform is a five-to-seven-year decision. It is not the place to be creative.
Layer two: The AI orchestration layer. This is where Cognigy, Google CCAI, Microsoft Azure OpenAI, or a purpose-built contact-centre AI platform sits. This layer owns the conversational AI, the intent understanding, the agentic orchestration, and the integration with knowledge bases and backend systems. This layer should be interoperable with the CCaaS platform — the major CCaaS vendors all have partnerships and certified integrations with the major AI platforms.
Layer three: The data and analytics layer. This is where interaction data is aggregated, analysed, and fed back into both the AI layer and the operations management function. Speech analytics, quality monitoring, performance analytics. This layer increasingly uses AI, but it is a different kind of AI from the interaction layer — it is retrospective and operational, not real-time and customer-facing.
The organisations that struggle architecturally are typically those that have tried to solve all three layers with a single vendor, or that have bolt-on integrations between products that were not designed to work together. The organisations that succeed have made deliberate choices about which vendor owns which layer, with clear integration contracts between them.
Where to start if you're in the 60%
BCG's roadmap for moving from laggard to scaling is sequential, not simultaneous.
First: get the governance right. One executive sponsor with a value mandate. Joint business-IT ownership of the programme. A clear target — specific, measurable, time-bound — not "improve customer experience" but "reduce cost-per-contact by 18% within 12 months through AI-assisted self-service."
Second: choose one workflow and pursue it to full deployment. In a contact centre, I recommend intelligent self-service for a high-volume, well-defined interaction type — whatever your equivalent of "balance enquiry" or "appointment scheduling" is. Not because it is the biggest value opportunity, but because it is the most achievable learning experience. A well-executed full deployment of a narrow workflow teaches you more about agentic AI in your environment than twelve pilots.
Third: redesign the downstream. Once the self-service layer is working, work backwards through the contact centre. Agent assist. Routing. Quality. Coaching. Each step builds on the data and operational learning from the previous one.
The future-built contact centre is not built in a single programme. It is built in a series of full deployments, each one extending the AI layer further into the operating model.
The five moves are not a checklist. They are an architecture — for the technology, for the organisation, and for the leadership commitment that makes both possible.
Data in this article draws on BCG's Build for the Future 2025 global study (September 2025), covering 1,250 senior executives across 25 sectors and 68 countries. Courtesy: Boston Consulting Group.