Blog

From data to action: can we trust AI with critical business decisions?

In an era where milliseconds can define the quality of customer relationships and your competitive positioning, the pressure to make faster, more contextualised decisions has never been greater. Enter AI – not just as a tool, but as a potential trusted agent and decision-maker. From forecasting demand to optimising supply chains and personalising customer experiences, AI is already influencing high-stakes choices. But as algorithms grow more autonomous and data more abundant, is it as easy as it sounds?

We continue our blog series building on the ideas of the 2025 Retail Innovation Forum in London by exploring the key innovation elements for closing the data-decision gap. Below are our top five themes:

  1.  Infrastructure choices matter: simplicity and ecosystem-readiness > customisation and developer-readiness

Technology architecture and data infrastructure decisions are increasingly important for organisations seeking to improve usability and relevance of their data-driven decision-making. Retailers and brands navigate a complex landscape of data sources which are individually sourced and channelled towards siloed user personas. While many stick to the status quo, unlocking the true value of ecosystem-driven innovation and AI efficiencies requires addressing this problem using data strategy and agile infrastructure.

Data strategy is increasingly important to streamline the path to efficiency and value from technology investments by defining what are the real needs of the users and what data assets are critical to achieve them. This bias for value is also contributing to a fundamental shift away from heavier ML-ready infrastructure towards simplicity and ecosystem-readiness, fixing other problems in the process: talent gaps (e.g. SQL > Spark experts are easier to find and train), DevOps drag (e.g. by choosing fully managed platforms), and cost transparency.

At an infrastructure level, it is important to explore beyond the obvious options – while many conversations today seem to debate ‘incumbents’ vs Snowflake vs Databricks, there is a broader world of emerging startups that make life even easier by solving extraction, storage, processing, and transformation problems simultaneously. And some go a step further, by not only piping data from many places into one, but by helping you curate which owned and leased data assets are relevant for different end user personas. As expected, this comes at a premium, so the need and usage scope depends on your use cases and level of agility you require for connecting across and beyond your data assets.

Your tech stack decisions must be made with the whole organisation and ecosystem in mind, otherwise you risk being left behind.

  1. Data governance and harmonisation is a prerequisite for turning it into insights

Alongside infrastructure decisions, preparedness for data integration and AI adoption is no simple task – it requires thoughtful data governance, standardisation (not only across sources but also across teams), and continued integration management. Privacy and security are key considerations for any data integration but many are not yet ready (e.g. 39% of organisations lacking a formal data governance framework according to TechRadar). Similarly, data quality is essential. If data is disorganised, incomplete, or inaccurate, even the most advanced AI tools cannot yield meaningful results. Building a clean, well-structured, and integrated data environment with a unified taxonomy ensures that organisations can trust the information they rely on for decision-making. Even a single type of data such as product information can become a major problem if not standardised and maintained under a unified taxonomy (e.g. a typical CPG without dedicated taxonomy science could have 1M+ product attributes across five to six internal teams, with information errors or gaps of ~20-50%). Data science experts like dunnhumby’s data consulting team have been solving these problems for 35+ years and their relevance in the new AI-enabled decision world is even more pronounced.

Data for the sake of data generates ‘broken’ insights unless it is harmonised.

  1. Integrating disparate data assets can unlock unprecedented insights and efficiencies

Collecting data is just the first step. The real value lies in combining data from multiple sources to create a comprehensive, actionable view of the business. For retailers and CPGs, this might include customer data, sales data, inventory levels, and market-level data. By integrating internal and external data sources, businesses can start to see patterns and correlations that were previously invisible. AI-powered analytics can help with integrating and enriching data assets to identify hidden patterns, predict future trends, and gain a deeper understanding of customer behaviour. This enables businesses to act with greater precision and confidence, ultimately driving better decisions and customer connections. This directly translates to better operational planning and improved efficiencies, which, in turn, leads to better customer satisfaction and sales (e.g. +28% forecast accuracy and -15% inventory cost reduction according to Gartner, 2024).

Most valuable insights are generated from breaking down data siloes, internally and externally.

  1. Turning data & insights into proactive, real-time action

Data in itself holds little value unless it can be quickly turned into action. For businesses to remain competitive, they must act on insights in real time. Companies using real-time data analytics are 23 times more likely to acquire customers, six times more likely to retain them, and 19 times more likely to be profitable (McKinsey).

By integrating AI into decision-making, businesses can respond proactively to emerging trends or challenges, giving them a significant edge in the market. What does this mean in practice? A streamlined workflow for decisions would incorporate innovation leaders start with using AI to identify most relevant signals and context factors that feed into the decision through simulation and forecasting. The next step is ensuring AI-driven insights are surfaced proactively and in real-time to the appropriate decision-maker who decides what is automated and what requires their input before action. Agentic AI can then support orchestrating action and machine learning closes the loop on measuring its effectiveness and future recommendations.

While this is an exciting transformation of data-driven decision making, most tools today still fall short at the hurdle of turning data into insights, let alone proactively recommending action and orchestrating it.

Insights for the sake of insights fall short of value creation without a contextualised path to action.

  1. Open innovation & cross-functional collaboration as a critical survival skill in the age of AI

A critical barrier to optimising decisions with data and AI is internal fragmentation. Often, data is siloed within different departments, preventing organisations from leveraging its full potential. By fostering cross-departmental collaboration and creating a unified approach to data, businesses can unlock deeper insights and drive more informed decision-making. In many cases, this involves starting cross-functional groups to break organisational siloes, uncover and prioritise AI use cases, and augment employee roles with new ways of working.

Collaboration also extends to external partnerships. Working with technology providers, data scientists, and AI specialists can help businesses overcome data-related challenges and accelerate innovation. External partners bring expertise, unique data assets, and fresh perspectives that can enrich a company’s data strategy. Furthermore, the journey to data-driven decision-making is never complete. Continuous improvement is key. Regularly reviewing data strategies, refining infrastructures, and integrating new AI capabilities will help organisations stay ahead of the competition and continue optimising their decision-making processes. Even the largest and most savvy technology organisations are embracing open innovation as a critical survival skill.

Research from AtScale emphasises that fostering a data-driven culture is less about deploying cutting-edge tech and more about developing a mindset where data-driven decision-making is the norm. This cultural shift encourages cross-departmental collaboration, leading to more informed decisions and better business outcomes.

Transformation of mindsets is the first step to AI-powered decision making.

 

So, can we entrust AI with critical business decisions? While the answer is different for each organisation and decision-maker, there are many foundational building blocks that can position you to start capture immediate value of innovation today. AI-powered decisioning is quickly transforming the pace of turning data into action. The first wave of use cases focused on innovating around machine learning-driven proactive recommendations based on existing data inputs, shortening the data to insight and insight to decision gaps. The second wave is currently incorporating agentic orchestration to close the gaps between insight to decision and decision to action. The next wave is focused on intelligence simulation and synthetic customers, marking a transformative shift from fixing ‘how’ we make decisions to ‘what’ those decisions should be.

 

Innovation insights powered by the dunnhumby Retail Innovation Network

These insights were collected by gathering feedback from our Retail Innovation Network, the fastest growing global open innovation programme for retail technology leaders. We would like to thank contributing panellists and showcase participants, including Marian Warm (dunnhumby), Daniel Palmer (Engine), Indranil Das (Microsoft), Mathieu Roche (ID5 Technology), Khaled Naim (Onfleet), Joshua Evans (VST), and Alex Dean (Snowplow).

To receive insightful thought leadership and engagement, explore joining the Retail Innovation Network, the fastest growing open innovation networks in retail, for free at www.dunnhumby.com/ventures and find out more about the next Retail Innovation Forum in Bentonville at www.dunnhumby.com/retail-innovation-forum.

The latest insights from our experts around the world

customer first data science analytics & machine learning services
Ready to get started?

Speak to a member of our team for more information

Contact us