Infocepts co-founder and CEO Shashank Garg argues that the central challenge in data and analytics is not a shortage of technology. The market expanded from roughly 60 or 70 tools to more than 1,400 in less than a decade, while the underlying business problems did not multiply at the same rate. Organisations now have many ways to solve the same problem but still struggle to translate analysis into recognised value.
Garg’s interview positions AI as an accelerator for that translation. It can discover patterns, explain why results change and let business users ask questions in natural language. Yet the strongest technology cannot repair poor data, weak engagement or a gap between analytics teams and decision-makers.
The Value Gap Is Organisational
Garg cited a survey in which 90 percent of data leaders believed their products created real business value, while only 39 percent of business leaders agreed. The difference suggests a communication and adoption problem rather than a purely technical failure. Analytics teams may deliver sophisticated work that does not connect clearly to decisions, timing or measurable outcomes.
Human obstacles include skills shortages, limited business engagement, resistance to change and weak data literacy. Data leaders need to explain work in the language of revenue, risk, service or operational performance. Business leaders, meanwhile, need enough data understanding to challenge assumptions and use insights appropriately.
Success criteria should be agreed before a project begins. When both groups define which decision will change and how impact will be measured, a dashboard or model has a clearer route to value.
Data Fluency Distributes Analytical Capability
Garg describes data fluency as the ability for employees across an organisation to understand, access and apply data as naturally as language. Specialised teams remain important, but value appears only when people use their output. Fluency can increase productivity and innovation while reducing dependence on a small central group for every question.
A “data concierge” can support this shift by improving access, trust and adoption through targeted interventions rather than building one enormous team. The approach treats enablement as a service: people receive help choosing sources, interpreting measures and moving from insight to action.
Fluency includes scepticism. Employees should understand data quality, uncertainty and the limits of a model. Democratisation without these habits can spread confident misinterpretation faster than centralised analytics ever could.
AI Can Reveal Questions Users Did Not Ask
Augmented analytics can explore why sales change or why market penetration differs by city, surfacing relationships a user did not know to request. Natural-language systems can translate a question such as sales by category and region into analysis across large datasets, then present results and possible follow-up insights in accessible language.
AI also supports data engineering through discovery, classification, metadata, governance and anonymisation. On infrastructure, models can assist with capacity prediction, anomaly detection, alert routing and root-cause identification. This makes AI an enabling layer across analytics rather than a separate vertical with no connection to business work.
Automated discovery still needs explanation. Users should see the data, assumptions and confidence behind a suggested driver. Correlation presented conversationally can sound causal, so experts must review claims before decisions depend on them.
Start With Outcomes and Controlled Experimentation
Garg advises an AI-first mindset but also encourages experimentation, explainability and continuing model training. Organisations can use hackathons or innovation contests to generate ideas, fail quickly and learn, provided promising prototypes move through more rigorous evaluation before production.
Immediate value can appear in retail monitoring, fraud controls, media recommendations or internal productivity. The correct starting point is the client objective, followed by assessment of data quality and the relationships that could influence the result. Technology selection comes after the decision and evidence are understood.
The interview’s durable message is that more tools will not close the business-analytics gap. AI can reduce grunt work and surface hidden patterns, while people create relevance, trust and action. Organisations that invest in fluency, shared outcomes and explainable systems are more likely to turn rapid technical change into dependable business advantage.
Technology portfolios should therefore be simplified around outcomes. When several platforms solve similar problems, maintaining all of them fragments expertise and increases cost. Leaders can retire legacy tools and select services that support the decisions employees actually make, rather than accumulating every new analytical feature.
Productised “Solutions as a Service” shift more responsibility toward providers by combining software, expertise and problem-solving. Customers gain speed and access to specialised capability, but they should retain visibility into methods and data. Outsourcing execution must not mean outsourcing understanding of consequential insights.
Model maintenance is another organisational capability. Market conditions and source data change, so a useful predictor can become obsolete without obvious technical failure. Monitoring, feedback and retraining schedules keep systems aligned with current decisions. Explainability helps users notice when an answer no longer fits reality.
The business-analytics gap closes through repeated collaboration. Analysts learn which constraints shape action, while business teams learn what the data can and cannot support. AI can make that conversation faster and more accessible. It cannot replace the shared language and accountability that turn an interesting pattern into a responsible decision.
That discipline also gives leaders a clearer basis for measuring whether an analytics investment changes behaviour, improves outcomes or merely produces more dashboards.
