X ![]()
As we approach the end of 2023, it is time again to study the critical issues facing organizations and how data and analytics can help leaders proactively address them in 2024. At this time last year, recession fears were building against a sharp rise in interest rates, inflation, supply chain disruptions, and market uncertainty.
Today, CEOs continue to grapple with big questions around enterprise-wide value from fast-evolving technology such as generative AI, recruitment and retention of qualified talent, building new products and services to stay ahead of the competition, and building organizational resilience while optimizing costs to deal with uncertainties.
A PWC survey shows roughly 75% of CEOs are focused on automation, upskilling, and deploying advanced technologies such as AI, with 60% doing it to reinvent their business for the future!
Infocepts' own State of Data & AI survey shows nearly 50% of enterprises are under-prepared to harness the full potential of AI, 25% continue to struggle with data quality issues, and 36% want to increase the usefulness of D&A in their organization.
Infocepts is bridging the gap between business and analytics for multiple retail, media, data syndication, financial, and health companies across the globe. Our 1,800+ associates help clients daily using solutions that address a range of business problems, such as connecting their D&A (data and analytics) strategy to business outcomes, building Data and AI products, or optimizing their investments.
Using our experience guiding leaders at multiple levels – from C-suite to operations – I share five D&A initiatives for 2024 to bring impact to your firm.
The ability of a business to generate and scale ANY value from AI (including generative AI) depends on how well it takes advantage of its data. Access to large quantities of high-quality data is a prerequisite and a top barrier to adopting AI.
While everyone wants to do the glamorous work of building and showcasing AI models, only some want to do the hard work of preparing and putting their data to work. Data leaders must take advantage of the AI attention in the Boardrooms to make a compelling case for instituting data enablement as a first-class citizen of their Data & AI strategy. What COVID-19 did for digital transformation, AI may do for data enablement!
You can start by forming a cross-functional data innovation team that includes business leaders, SMEs, data scientists and technology leaders to brainstorm unique value-driven business use cases and develop data maps needed to realize them. Then, create a comprehensive inventory of both your structured & unstructured data, metadata, and lineage with a plan to close the gaps.
You should also anticipate data-related issues related to domain expertise, education, accountability, or business workflows that may cause compounding downstream impact on your AI solutions & address them proactively. Finally, take a high-value use case and partner with a business leader and your CFO to make the case for data enablement. Infocepts Insights-IQ supports such requirements.
AI is fast becoming a general-purpose technology, much like electricity, with many uses. Simply defined, it is a form of prediction made using data that enables judgments to improve actions that humans or machines must take to achieve intended outcomes. It can be as simple as predicting the next word in an email, generating a few lines of code, or recommending the next best action for a given situation!
Companies that integrate automated AI in redesigned business workflows are more likely to win in the long term over companies that use AI for incremental improvements. However, AI projects have a high failure rate and take a long time to move from pilot into production.
There are many barriers to success, including lack of data, incorrect framing of problems, lack of trust in AI, and inability to deal with human aspects of integrating AI into production. Additionally, clients struggle to decide – rightly so – which technology to use when there are 100s of niche technologies out there, most appearing to do similar things. To seize the opportunity and reduce risks, establish a formal AI Labs program for your business.
To remove the barriers, you must frame your hypothesis to clarify what you are trying to predict, what outcomes you want to achieve, and what actions must be taken based on those predictions and by whom. You then need to acquire the data, build and evaluate the model, and deploy it into production for implementation using the right technology.
There are two keys to success:
build a responsible and repeatable process for experimentation & scaling, and
create a system of experimentation and keep it separate from your system of implementation.
This ensures you can quickly experiment with fast-evolving technology without lock-in or buyer's remorse. Infocepts DiscoverYai is one such fully managed AI Labs solution that has proven to help businesses effectively.
The public introduction of Open AI's ChatGPT 3.5 in November 2022 and subsequent developments such as Google (PaLM2), Anthropic (Claude), Meta (LlaMa) & GPT-4 has democratized Generative AI. Knowledge workers in every field – from doctors to developers to economists – are using it and wondering what this means for their field and roles.
A recent St. Louis Federal Reserve Bank study demonstrated that conditional inflation forecasts generated using Google's PaLM are more accurate, cheaper, and faster to produce than the Survey of Professional Forecasters! Meanwhile, many CEOs still wonder if this is hype or a game-changing technology for their business.
Data leaders will play a pivotal role in helping their businesses navigate and operationalize their generative AI strategy. In 2024, organizations must move from discussing use cases to successfully experimenting and implementing them.
Generative AI will produce value in three areas:
increase developer productivity,
improve how customers consume your goods and services while helping launch new ones, and
improve business operations.
Examples include using co-pilot tools to reduce the time and effort to convert data into insights, offering interpretive services such as helping your customers understand their bills, and enabling your procurement to reduce the time for market research dramatically.
Generative AI projects include unique cost drivers depending on how you adopt LLMs and implement business apps. To adopt an LLM, you can "Talk" to a model such as ChatGPT-4 using prompts. This approach has quick time to value and low cost but does not offer any moat for your business.
You can "Tune" existing models by bringing "your data to the models" (e.g., Azure Open AI or Amazon Bedrock) or the "models to your data" (e.g., Cohere). This approach has a short time to value, medium cost, and low moat.
Finally, you can "Train" a model if you have proprietary data supporting your business. This approach has a long time to value, high cost, and high moat.
When implementing apps, remember that only 10% of your effort will go on modeling, 20% on data and tech, and 70% on quality, change, risk, and workflow management. Because AI outcomes are probabilistic, realizing business value requires significantly more effort to integrate its results into business workflows.
A 2023 D&A Leadership Executive survey shows that while 92% of data leaders say that their data products are delivering business value, only 39% of business leaders agree. There are many reasons why this gap continues to remain despite evolutionary iterations in how value is generated from data, such as treating data as a product or embracing data fabrics.
One reason is that IT leaders limit their role to the provisioning of secured technology and data and rely on business analysts to use self-service to generate insights for business leaders.
Infocepts finds the generally accepted industry definition of a data product (purpose-built data sets with metadata, interfaces, access control, versions, lineage) limiting since it only targets the data layer. For data to create business value, it must be applied systematically using full-stack domain-driven data products.
A full-stack data product clearly specifies which personas it addresses with a focus on the end users of insights and not those responsible for generating them.
Secondly, it handles business use cases relevant to the chosen domain with consistent design patterns to achieve business outcomes – the missing link.
Thirdly, it also includes a highly intuitive application interface to support the actionable needs of the personas.
Lastly, the product consists of a unified data layer with modeled data and supporting metadata. The entire stack is managed using versions so that the product can be built iteratively and support evolving use cases & growing data.
Infocepts Employee360 is one such fully managed data product for employee analytics.
The Modern Data Stack promised to make D&A accessible, flexible, faster, easier, and cheaper than the traditional data stack. In the past five years, there's been a 10X growth in D&A technologies. Modern enterprise environments routinely consist of 25+ technologies with several redundancies.
For example, one of our clients is moving from Netezza to Snowflake using Databricks as an orchestration layer!
Clients get into these situations due to the need for modernization, aggressive selling by technology vendors, perceived niche requirements, and M&A actions. The result is that shadow IT is prevalent, complexity has increased, talent problems have worsened, portfolios have bloated, and costs have spiraled.
Infocepts advocates for D&A portfolio rationalization as a pivotal tactic to ensure that you save & redirect money in 2024.
A high-level application rationalization process includes identifying business needs, assessing business value, technical fit and total cost of ownership, making recommendations, and implementing and monitoring the benefits.
Its execution challenges include a lack of engagement from stakeholders who possess the knowledge but may have a personal stake in maintaining the status quo, subjective assessments and complex TCO calculations, lack of fractional skills needed to execute, alignment on what's suitable for the enterprise, and change management.
For successful rationalization, Infocepts recommends data leaders maintain minimum information about every asset that costs you money. This includes data, technology, tools, resources, and infrastructure. Consider outsourced services to ensure you have access to the technical, business, and management skills needed for rationalization independent of your prevailing priorities.
Lastly, partner with the CFO – they are best suited to understand what the CEO values, what business leaders need to deliver for the company, what constraints are driving decisions, and keep emotions out of decision-making.
Infocepts AirCover is one such fully managed service to help clients rationalize their D&A choices using data-driven decisions supported by cost & value.
About the Author:
Subhash focuses on generating evidence-based insights, developing innovations, and providing advisory services to improve business outcomes. He leads Infocepts' Global Practices & Delivery organization to develop solutions, strengthen competencies, oversee delivery & fulfilment, and develop people. Subhash brings two decades of experience in mission-critical applications, data & analytics solutions, and large-scale managed services.