
State of Analytics in the Age of AI: Six Key Takeaways for Data and Analytics Leaders
AI can now generate a polished dashboard in seconds. But how do we know whether the numbers are reliable, what they mean, and whether they can support a business decision?
These questions were at the heart of State of Analytics in the Age of AI, a virtual summit held on 22 October 2026.
Curated and chaired by Ben Clinch of Clinch Data Services, the event brought together Simone Steel of Deutsche Bank Corporate Bank, Dr Herbert Prickarz of Bosch Mobility, Johanna Kruckenberg-Turen, formerly of H&M Group and Handelsbanken, and Glen Armashula of Accton.
Their discussions explored how AI is changing analytics, from the way business users ask questions to the governance, skills and organisational structures needed to make the answers trustworthy.
Each speaker brought a different perspective, but several ideas kept coming up throughout the summit. We’ve brought them together in six practical takeaways for organisations thinking about what comes next for analytics.
1. Start with the decision, not the dashboard
Analytics teams have traditionally asked business stakeholders what data or reports they need. But this question already assumes that the stakeholder knows what information will help them.
During the panel discussion, Simone Steel challenged this approach. What if we started by asking which decisions people want to make?
A business leader might request a sales dashboard, for example, when the underlying question is whether to enter a new market or change an existing strategy. Understanding that decision first gives analytics teams a clearer starting point for identifying the evidence required.
Dr Herbert Prickarz shared a practical example from Bosch. During a management meeting, he asked participants to identify the data they needed to understand and steer their business. Initially, the requests varied considerably. As the discussion progressed, managers began questioning what information they genuinely needed.
The distinction becomes even more important as AI changes how analytics is used.
In her keynote, Simone described a future in which analytical outputs increasingly feed directly into business operations. Instead of presenting a recommendation for someone to consider, an AI-enabled process might act on that information, for example by adjusting a customer interaction or triggering an operational response.
This creates new requirements for understanding where information comes from, how it is interpreted and what happens when it is used.
For analytics leaders, a useful starting point is to establish three things before commissioning new analytical work: the decision being supported, the evidence required and the action that will follow.
2. Give everyone a shared map of the data landscape
In large organisations, different teams often operate with different views of the same business.
Technology specialists understand systems and architectures. Business leaders understand customers, operations and commercial priorities. Data professionals have their own models and terminology. Bringing these perspectives together can be difficult, particularly when they are expressed in different languages.
Simone Steel proposed a compelling way to bridge these gaps: a shared map of the organisation’s data landscape.
Her inspiration came from the way people navigate physical environments. Maps make complex surroundings understandable through familiar boundaries, symbols and layers. They help people see where things are, how they connect and what lies beyond their immediate area of interest.
The same approach could help organisations understand their data.
Simone brought this idea to life with a simple example. Imagine two teams at the same bank: one wants to improve customer service, while the other needs better information about financial and market risks. Both are looking to invest in analytics to achieve their goals.
Although the teams had different goals, bringing their information needs together on a shared map helped reveal potential overlaps, particularly around customer data and external market signals.
Instead of treating the two initiatives separately, the bank could identify shared data needs and opportunities to invest in common capabilities.
A shared map can also help teams spot gaps, assess the impact of proposed changes and better understand how business functions and data assets are connected.
Keeping that map up to date is equally important. As Simone pointed out, a static map would quickly become outdated. It needs to evolve alongside the organisation’s systems, processes and information, with metadata that captures how data is connected and used.
Ultimately, the value of a shared map lies in giving people a common understanding of the business. It provides a foundation for discussing investments, managing dependencies and making informed decisions across teams and departments.
3. Make the difference between exploratory and trusted analytics explicit
A question as simple as “How many people work for this company?” can produce different answers depending on who is asked.
As Herbert explained during the panel, HR and controlling departments may calculate workforce numbers differently. One definition might include contractors, while another excludes them for financial reporting purposes.
Both figures may be valid within their respective contexts. Problems arise when people use them without understanding the difference.
AI can make these differences harder to spot. While an AI assistant can produce an answer in seconds, the definitions, calculation logic and sources behind that answer may remain unclear.
This is where the distinction Glen Armashula made during the summit’s opening conversation becomes particularly useful. He described two types of analytics: exploratory and verified.
Exploratory analytics allows people to investigate questions, test ideas and identify patterns quickly. Flexibility and speed are valuable here, and the results may require further validation.
Verified analytics supports recurring business needs and formal decisions. It requires agreed definitions, clear ownership and the ability to trace results back to their underlying data.
Some exploratory questions come up again and again. When they do, it may be worth turning the answer into a verified, reusable analytical asset.
Johanna Kruckenberg-Turen took this idea a step further, applying it to AI-generated content. She suggested treating AI outputs as exploratory by default. Before they can be used to support formal decisions, they need to be grounded in trusted metrics, with appropriate controls in place.
That doesn’t mean every department needs to use the same definitions. As Herbert pointed out, differences may be necessary to meet regulatory requirements or serve different business purposes.
What matters is making those differences clear.
Simone also stressed that statistical rigour must be accompanied by transparency about the limitations of the underlying data. Even when data has been quality-assured, users need to understand what those assurances actually guarantee.
For organisations bringing AI into analytics, trusted business definitions are a critical foundation. AI systems need access to the same agreed meanings and reliable metrics that people use to interpret business performance.
4. Stop report proliferation before AI makes it worse
The ability to create dashboards almost instantly sounds appealing. It also raises an uncomfortable question: what happens when everyone can generate a new dashboard whenever they want?
Simone drew a parallel with the proliferation of spreadsheets in financial services.
Spreadsheets gave business users enormous flexibility. Over time, however, they became central to critical financial processes, including the development and use of statistical and risk models. With inconsistent calculations and limited oversight, banks often struggled to get a clear and consistent picture of their financial relationships with one another. This became especially problematic during the financial crisis.
AI-generated dashboards could create similar difficulties on a much larger scale.
A visually convincing report may contain incorrect figures, questionable assumptions or definitions that conflict with established business standards. The problem is that AI may not recognise the difference between a correct answer and a plausible approximation.
As the number of reports grows, understanding which ones people actually use becomes increasingly difficult.
Johanna offered a surprisingly practical response: start retiring unused reports.
The process starts with understanding how existing reports are actually used. By looking at usage data, organisations can identify reports that haven’t been opened for 30, 60 or 90 days and select potential candidates for retirement. The next step is to communicate the plan to stakeholders and gather their feedback.
These conversations can reveal who still uses each report, which business decisions depend on it and who should be responsible for maintaining it.
With a clearer picture of their actual value, organisations can decide which reports should be retained or improved and which can be archived. This helps reduce unnecessary maintenance and free up analytical capacity.
Johanna shared examples of organisations retiring hundreds, and sometimes thousands, of reports through this approach. The process was often less disruptive than stakeholders initially feared. It also helped identify which analytical assets deserved closer governance.
As AI accelerates report creation, organisations can focus their governance efforts only on the metrics and reports that support important business decisions.
There is also an opportunity to rethink how AI tools create analytics in the first place. If AI agents can consult existing reports, approved definitions and reusable calculations before generating new outputs, they may help reduce duplication rather than add to it.
5. Treat governance as something people do every day
Most organisations already have governance policies, standards and procedures in place. At the same time, inconsistent definitions, duplicated work and unclear ownership remain common challenges.
Johanna’s keynote explored why governance so often struggles to deliver results in practice.
Her central argument was that governance is social infrastructure. Its effectiveness depends on relationships, shared ownership and everyday behaviour.
Policies and technical controls provide structure, but people need to understand, accept and apply them.
Alongside report retirement, Johanna shared two approaches she has used to make governance work across organisations.
The first is to build communities around people who already share knowledge and support their colleagues. These natural ambassadors can help establish common practices, spot problems early and encourage others to get involved.
Active participation is what makes these communities valuable. A small group of engaged members can have a greater impact than hundreds of people who rarely contribute.
Her second approach involves creating working groups with a clear purpose and specific goals. To be effective, these groups need a well-defined problem, a mandate, an accountable leader, a deadline and concrete deliverables. They should bring together people responsible for making decisions and those affected by them.
This structure helps turn governance discussions into agreed standards and practical changes.
These responsibilities increasingly extend to AI-generated analytical outputs. Organisations need to record who created them, what data was used and which prompts informed the results.
The same governance principles still apply: clear ownership, accountability and shared understanding. As AI-generated analytics become more common, it becomes increasingly important to apply these principles consistently across the business.
6. Develop people’s judgement alongside their AI capabilities
As AI becomes more capable, the skills organisations need from employees and leaders are changing.
Dr Herbert Prickarz’s keynote focused on the people responsible for making data, analytics and AI useful within organisations.
At Bosch Mobility, data and AI learning programmes help people build a shared foundation in areas such as data management and governance. To make the training relevant to people’s everyday work, the programmes include additional content tailored to individual divisions and business functions.
Herbert also highlighted the value of connecting learning with real business projects. Participants should start applying their new skills while they are still learning and, where possible, demonstrate the business value of what they have learned.
Bosch keeps improving its learning programmes by tracking how many employees take part, how relevant they find the training and what they share in their feedback.
The company has also opened data stewardship training to people in other roles, including analysts and data scientists. This has helped build a broader community where colleagues from different disciplines can learn from one another.
But technical competence is only part of the challenge.
Herbert explained how leadership needs to evolve as people increasingly work alongside AI agents. Leaders need to become better at framing problems, setting clear boundaries for decisions and coordinating work between people and automated systems.
They also need to understand their organisation’s data and processes well enough to assess AI outputs and remain accountable for decisions.
The panel also explored an interesting question: could getting instant answers from AI make people less inclined to think critically?
Simone raised concerns about cognitive fatigue and the risk of becoming too dependent on AI answers that sound confident and are always readily available. Herbert offered a different perspective, suggesting that AI could help people tackle more complex problems than they could manage before.
The impact of AI on human judgement is still being explored, and these questions remain open.
One thing is clear: organisations need to help people develop the ability to ask meaningful questions, evaluate evidence and challenge answers when necessary.
Glen linked this discussion to four familiar barriers to analytics adoption: motivation, skills, trust and user experience.
AI can make analytics easier and more enjoyable to use. Organisations still have an important role in building the skills, motivation and trust needed for successful adoption.
Where should analytics leaders go from here?
Throughout the summit, speakers kept returning to challenges that data and analytics professionals have faced for years: inconsistent definitions, fragmented information, unclear ownership and the difficulty of turning governance principles into everyday practice.
AI makes addressing these challenges increasingly urgent.
In his closing reflections, Glen Armashula summed up the discussion with three priorities:
- Give everyone a shared map of the organisation’s data landscape.
- Make sure someone owns what is on that map, including the information and definitions people rely on.
- Invest in the people who will use it, developing the skills, understanding and judgement needed to make good decisions.
These priorities are closely connected.
A shared map helps people understand what information is available and how it connects. Clear ownership helps establish what that information means and who is responsible for maintaining it. Investing in people gives them the skills and judgement to turn that understanding into better decisions.
Putting these principles into practice can start small.
Choose one important business question that comes up regularly. Identify the data and metrics behind the answer, check whether their definitions are agreed, establish who owns them and assess whether the resulting analytics can be trusted for the decisions they support.
This exercise alone may reveal opportunities to improve governance, reduce duplicated work or make existing knowledge more accessible to AI.
Glen ended the summit with a practical challenge inspired by Johanna’s keynote: choose one idea and start putting it into practice the following week. Begin with a small experiment, learn from it and build on the results.
As AI changes how analytics is produced and used, organisations have an opportunity to build on the knowledge, analytical assets and experience they have already accumulated.
The decisions they make about that foundation today will shape how much value they can realise from AI in the years ahead.
Putting these ideas into practice
Many of the summit’s recommendations start with a clear understanding of an organisation’s existing analytics. Accton helps put these ideas into practice by bringing reports, usage, ownership and business definitions into one analytics catalogue and creating one shared business context layer. This gives organisations a stronger foundation for better governance, more effective use of their analytics and reliable AI-powered insights.
Explore the summit sessions
The summit’s keynote presentations and panel discussion offer more practical examples, expert insights and ideas to help data and analytics leaders address these challenges.
Watch the State of Analytics in the Age of AI session recordings — https://accton.trumpet.app/live/m3wDeoVqBk/state-of-analytics-2026-10






