conan saha

In AI projects, it’s easy to get distracted by what’s technically impressive. But when you’re in transformation—especially in something as operationally intense as Healthcare BPO—you need to stay laser-focused on what actually moves the needle. One approach I’ve found useful is a “problem-first” mindset. That means, before we talk about models or data pipelines, we start with a basic question: What real business issue are we trying to solve—and how will we know if it worked? That keeps things grounded. Let’s say a team wants to use AI to summarize provider-patient calls. Interesting idea, but I’d ask—what are we solving here? If agents are spending too long on post-call documentation, then sure, maybe it helps reduce AHT or frees up bandwidth for quality. But unless there's a clear outcome—less handle time, fewer compliance flags, better throughput—it’s just tech for the sake of tech. That’s where something like COPQ (Cost of Poor Quality) or even a basic cost-benefit check can help. In one project, we looked at automating the review of denied medical claims. But before jumping in, we estimated how much rework was being caused by misrouted or poorly coded denials—turned out to be in the millions. That changed the conversation completely. Now we weren’t just doing AI—we were solving a high-cost operational problem. So the trick, really, is to flip the usual sequence. Do not ask, “What can we automate with AI?” Instead ask, “What’s costing us money, time, or customer trust—and can AI help fix that?” It’s a mindset that helps filter the noise. Because at the end of the day, nobody’s impressed by cool tech that doesn’t actually deliver business value.

Change management often makes the difference between a DMAIC project that sticks — and one that fizzles out. It’s not just a final step after solutions are rolled out. It actually plays a role in every single phase of the DMAIC cycle. If you ignore that, you’ll usually see great ideas fall flat. Define – Setting the Stage Early At this stage, change management is all about creating that shared understanding: why are we doing this? why now? You want stakeholders to feel the urgency and know they’re part of the solution from day one. In one of our telecom process improvement projects, we were trying to cut down Average Handling Time. We looped in team leads right at the start. They were worried it might burn out agents, so we adjusted the goal slightly and added stress management checkpoints. That early conversation saved us from a lot of pushback later. Measure – Getting Everyone to Trust the Numbers Here, people start getting skeptical. “Where’s this data from? Is it accurate?” Change management here is less flashy, but super important — it’s about building confidence in what we’re measuring. In a healthcare case, some nurses didn’t buy into the data about discharge delays. We didn’t fight it — instead, we got them involved in mapping the current workflow and verifying how data was captured. Their buy-in increased dramatically just because we treated them like co-owners of the truth. Analyze – Digging into the Root Causes (Even the Uncomfortable Ones) This is where change management helps navigate politics. You’ll find root causes people don’t want to admit — outdated policies, favoritism, or even team-level issues. I remember a retail returns project — root cause traced back to how sales reps at the POS entered product details. The data exposed some clear gaps, but instead of blaming, we framed it as a training and system fix. That reframing made the solution more acceptable to everyone. Improve – Making People Actually Try the New Stuff Even the best solutions won’t land if people don’t use them. This is where you need proper rollout plans, pilots, training — and just listening. You’ve got to make it feel like their solution. For an utilities account, we moved to a pull-based inventory setup. Initially, warehouse staff were hesitant — they thought it meant more manual work. So, we ran a pilot in just one location, added some quick visual aids, and celebrated small wins. This helped convert skeptics into supporters. Control – Keeping the Gains from Fading This is the most ignored phase for change management — and ironically, the most critical for sustaining impact. People slowly slip back into old habits unless you lock the new ones in. In a BFSI complaint-resolution improvement we worked on, we tied the new metrics into their existing weekly ops reviews. We also added recognition for teams that hit zero-complaint weeks. Just those two tweaks kept the momentum going well after the project closed. Role of an MBB in Driving and Sustaining Change A Master Black Belt can’t just be a Six Sigma “tools” expert. Their job is to be a strategic influencer and coach. I’ve found the most effective MBBs don’t just push DMAIC steps — they spend time understanding people, motivations, fears. They help build the story early — connecting the project to broader business needs. They teach GBs/BBs how to manage resistance, not just do root cause trees. They embed rituals and systems that keep improvements alive beyond project closure.

In order to ensure the AI system evolves with the process, I would design the AI system with a built-in feedback mechanism that’s linked to the human lead process improvements. Every time a team completes a PDCA, DMAIC, or A3 cycle, the key insights and process adjustments should be captured and fed back into the AI system. It would be a standard process to document every process improvement initiative in a specific template, generate relevant training data and feed this to the AI. An MBBs role here would be to bridge the gap between the process excellence teams and the technical teams managing the AI. The MBB will also be responsible to create the templates needed to capture the insights and updated process steps (If applicable) which can be feed back to the AI through the established feedback mechanism. The MBB would also need work with the technical team to set up a process to effectively use this template for training data generation. Additionally creating and managing a governance plan should also be the MBBs responsibility.

Below listed are the traits of a good AI Governance Model – 1.) Clear and Fair – Identify and reduce any biases in the model, provisioning options for human interventions for big decisions and ensuring POCs are defined and known to all stakeholders in case something goes wrong. 2.) Strategic Fit – Ensure every AI deployment directly supports the organization's strategic goals and delivers clear value 3.) Managed Lifecycles: AI systems should have a defined route from initial conception to continuing upkeep. This calls for extensive testing prior to deployment, ongoing performance monitoring, and a defined procedure for modifications or even retirement. We require accurate documentation of everything. 4.) Training – Ensuring that relevant teams are well trained to work effectively with the AI. Also, teams need to be aware of what the AI can and cannot do. Stakeholders of an ideal AI Governance Model – 1.) Leadership 2.) AI oversight group 3.) Ethics and compliance team 4.) Internal Auditors Mechanism to ensure both agility and control 1.) Smart Risk Assessment – Design and approval framework aligned with the risk quotient of the deployment. 2.) Use of standardized tools and reusable modules – Provision pre-approved tools and use of reusable building blocks to cut down redundant work. 3.) Build in governance from day 1. 4.) Centralized guidance from the AI oversight team. 5.) Clear and defined RACI. 6.) Provision Human-AI collaboration for high risk decisions.

AI adoption is a key focus area for most companies today if not all. In this scenario it has become very common that MBBs and AI Solution Architects often end up working on the same transformation initiative. Although both the teams have different approach to the same problem, here are a few strategies which MBBs can follow to ensure that AI initiatives are aligned with process excellence, customer value, and organizational priorities – 1.) Business Case – Ensure every AI initiative is has a clearly defined business problem and is aligned with the strategic objectives of the organization. 2.) Embed Process Excellence Methodologies – a. Apply established methodologies like DMAIC (Define, Measure, Analyze, Improve, Control) or DMADV (Define, Measure, Analyze, Design, Verify) to the AI development lifecycle. b. Utilize Value Stream Mapping (VSM) to identify areas of the process where AI can be applied. 3.) Customer Journey Mapping - Create a map of the customer experience and pinpoint the precise points of contact where AI may enhance communication, lower barriers, or provide tailored value. 4.) Establish a robust governance and collaboration framework with shared goals regular review/tollgates 5.) Benefits Realization Management- Establish a structured procedure to monitor and document the true advantages of AI projects, guaranteeing responsibility and proving their worth.

Let’s take an example of a healthcare process which handles provider enquiries. Traditionally, Lean Six Sigma have been used to optimize this process. For example, use of DMAIC Projects to reduce AHT or improve FCR. However, in today’s highly dynamic contact center environment Lean Six Sigma methods might face some limitations – 1.) Lean Six Sigma primarily focus on incremental improvements and are not designed for radical reimagination or innovation. 2.) Reliance on historical data – This makes it mostly a reactive intervention and not best suited for proactive identification of potential issues. 3.) Limited by human centric bottlenecks like - a. Limited speed of processing information b. Scalability challenges c. Cognitive load d. Human Bias Here are some examples of AI based interventions for the given process – 1.) AI powered self-service portal – Handing provider queries in natural language (voice or text) and provide immediate resolution like claim status, Eligibility Enquires and Pre Auth requirements. 2.) Intelligent Virtual Assistance – To handle routine and simple calls. 3.) Assisted Call Handling – Provide ‘next best action’ prompts, pull relevant provider history and auto populate documentation 4.) Realtime customer emotion and sentiment analysis and live feedback with relevant scripts to guide the agents. 5.) AI powered knowledge bases for easy retrieval of information and customized capsule training modules based on agent interactions.

In my projects I typically leverage the effort impact matrix to pick the best ideas to fix a given problem. Based on the scenario adding a criticality element to the effort impact matrix might be recommended. For the given example, using a cost benefit analysis will be a better idea since it allows for a more direct financial comparison of large investments like additional headcount vs investing in better tools and effectively quantify ROI. Below are some of the steps which I would take in this example to avoid resistance to change – 1.) Effective and transparent communication 2.) Stakeholder involvement since early stages of analysis 3.) Utilize A/B testing to ensure corrective actions can be promptly taken in case the interventions fails.

Below are some of key steps which can be taken to ensure real causes and being identified and pursued in the Analyze phase – 1.) Use of structured data – It is very important to use a structured, data driven approach and to avoid relying on anecdotes. 2.) Ensure the problem statement is specific and quantifiable. 3.) Ensure proper data stratification. 4.) Follow 5 Why effectively – Do not stop at the first why, keep digging until a proper root cause is revealed.

Employees being the frontline users in any given process, Voice of Employee (VoE) is a very useful tool for the Business Excellence Team. Here are some of the key areas where it can help – 1.) Identifying Process Improvement Opportunities – Employees have a higher visibility of broken processes, system issues. Additionally, overtime employees develop best practices based on their exposure to the process which can benefit the entire process if successfully replicated. 2.) Employee Engagement and Retention – Ensuring employees are heard and their suggestions are evaluated for implementation in a transparent manner, can improve employee engagement and retention. Challenges in gathering VoE – 1.) Lack of structured VoE framework 2.) Lack of trust and fear or retaliation among employees 3.) Failure to act on the suggestions 4.) Effectiveness of action items – It can be challenging to identify actionable insights from the VoE data 5.) Resistance from management (supervisors, managers, etc) - Ensuring the feedback is viewed as an improvement opportunity vs criticism by all leadership stakeholders. Best Strategies – 1.) Create a structured framework in place with a clear RACI for all stakeholders and a proper governance structure. 2.) Building a culture of trust, ensuring anonymity of the surveys, promoting leadership buy-in and commitment. 3.) Publishing action items and implementation status with the employees 4.) Manager training and accountability Example – Here is an example from one of the previous organizations – A VoE program was designed for an 300 FTE process to identify employee pain points. Feedback was collected and a structured manner, action items were created and progress of implementation was regularly published, below was are the benefits which were realized – 1.) 26 validated process improvement ideas (Themes – AHT improvement, Quality Improvement, Automation) 2.) 20% reduction in EWS 3.) 8% reduction in attrition

Absolutely, Key Risk Indicators (KRIs) can definitely be used to manage a process. However, their implementation often faces challenges that lead companies to prioritize Key Performance Indicators (KPIs). One core issue with KRIs is the difficulty in defining and consistently tracking them. Identifying truly predictive indicators, rather than just historical data, is complex. Plus, KRIs can sometimes trigger 'false positives', leading to wasted effort. KPIs, conversely, are typically based on clearly defined, agreed-upon goals, making them generally easier to quantify and report. KPIs also offer a more immediate view of current performance, enabling quicker corrective actions. This direct, performance-focused feedback loop is often why organizations favor KPIs for daily oversight. Despite these challenges, the benefits of integrating KRIs are significant: Benefits of Using KRIs 1.) Proactive Risk Management: KRIs shift you from reacting to problems to preventing them. They signal potential issues early, allowing you to address risks before they escalate. Example: A KRI showing the increasing age of IT infrastructure can flag risk of system failures, prompting proactive upgrades and preventing costly downtime. 2.) Early Warning System: They provide timely alerts about rising risk exposure, enabling swift intervention and mitigation. Example: A KRI tracking the percentage of loans overdue by 30+ days in a financial institution offers an early warning of credit risk, prompting a review of lending policies. 3.) Improved Decision-Making: KRIs offer data-driven insights into potential threats, helping leaders make smarter decisions about resource allocation and strategic planning. Example: If a KRI reveals increased employee turnover in a department, management can investigate root causes like workload or management style, then make informed decisions to improve retention. Limitations of Using KRIs 1.) Defining Relevant KRIs is Complex: It's tough to identify truly predictive KRIs. Many end up being lagging indicators that tell you what's already happened, not what's about to happen. Example: "Number of security incidents" is lagging. "Number of unpatched critical vulnerabilities" is a more predictive KRI for future risk. 2.) False Positives/Noise: Some KRIs might trigger alerts that don't indicate a real risk, leading to unnecessary investigations and 'alarm fatigue'. Example: A KRI for "employee login failures" might spike due to a temporary network glitch, not a malicious attack, causing overreaction. 3.) Data Collection & Accuracy Challenges: Effective KRIs demand reliable and timely data. Issues with data sourcing, quality, or integration across systems can hinder their effectiveness. Example: If a KRI relies on manual data input, inconsistencies or delays can make the KRI unreliable. 4.) Resource Intensive: Implementing and continuously monitoring a comprehensive KRI framework requires significant resources, including technology, personnel, and analytical capabilities. Example: Building automated KRI dashboards for numerous processes demands substantial investment in IT infrastructure and data analytics expertise.