AI offers a huge opportunity to the companies willing to embrace it. And there’s some urgency: 84% say they won’t achieve their growth objectives, and three-quarters of executives believe they risk going out of business in five years unless they scale AI.
The stakes are high. But rushing to scale new ideas quickly requires a well-constructed data strategy – which we define as a design and intent that underpins what data is being captured, in what way and for what purpose. Like a house built on weak foundations, an AI solution built on weak data with no solid strategy may deliver some near-term value but doesn’t stand a chance at scale or delivering results in the long term. The data strategy drives value as much as AI does.
There are five key questions that senior leaders should be asking themselves when building their data strategies to help ensure AI gets to production and is ultimately scaled successfully. These critical questions will help guide leaders to build a data strategy based on culture, data quality, and privacy, to unlock the value potential of AI.
1. How do we develop a data-driven culture?
Building a data-driven culture begins with buy-in from the top. Senior leaders need to show employees what is possible with data and invest in the tools and resources that’ll empower their employees to achieve those possibilities. They need to communicate the benefits of working with data and encourage behavioral changes by simplifying and incentivizing the use of data-driven insights in making business decisions.
One company we worked with had invested significantly in business intelligence and analytics capabilities for insurance underwriting but struggled to have their employees use them. To help, we developed an interactive prototype of priority use cases that reflected how users worked and behaved. We embedded insights into the flow of work so employees could – at a glance – see how effective performance was. We also provided an analytics-driven view of upcoming activities, which users could refine in a simple manner. And we created simple underwriting checklists and flags for important, but easily forgettable aspects of the process.
By developing tools that reflect how people naturally behave and use data on a daily basis, we can begin to build a culture that organically embraces data and analytics. But having the tools and incentives is only one piece of the puzzle. Extensive, ongoing training is essential in building data literacy, as is hiring the right people.
2. How can we trust the quality of our data?
For starters, when talking about ‘data quality’, we look at a combination of factors such as completeness, accuracy, lack of bias, relevance and timeliness of data in relation to the insights we are trying to generate. And a lot hinges upon businesses having high-quality data. For one: adoption. If an employee or business group’s willingness to use data is contingent upon their trust in the data, then it’s going to be important to build confidence in the quality of the data to build that trust and encourage usage.
So, how can you measure and maintain data quality?
First, companies should establish effective data-quality processes and frameworks around storage, management, and transfer. Designated data owners must act as custodians of data quality in each domain.
A company can perform ad-hoc, manual spot checks on each data attribute (e.g. recency), but that can be inefficient and incomplete. Alternatively there are solutions like Accenture’s Data Veracity Offering, which help evaluate the provenance, context and integrity of data on a systematic basis. Pre-built tools and frameworks are used to determine the quality, risk and relevance of data and produce a data veracity score to quantify the quality of the data and track improvements over time.
It’s this combination of clear ownership and accountability, along with solutions to measure data veracity that can help prove data quality and establish trust with business users.
3. How do we harness innovation in our data platforms?
While culture and data quality are important to building a solid data strategy, platform innovation is essential to future-proof that strategy. By bringing new sources of data, diversifying underlying technologies and applying new technical approaches, you can deliver much sharper, near real-time insights across the enterprise.
Where can companies find new data sources to feed this innovation? We recommend tapping into unstructured data sources that may have been untouched until now. For example, within legal limits, manufacturers can use camera images and video to assess quality and functionality of what’s being manufactured on the production line.
Or, if companies have already exhausted the valuable data sources within their walls, they can look to bring in trusted, third-party data to complement or fine-tune the insights they already have. They can also look to incorporate data from the edge, whose real-time analysis of sensor data can help, for example, predict maintenance for a variety of industrial and energy production equipment to improve operations.
Then, think about the ingestion and management of that data. Five to 10 years ago, most organizations used batch processing through extract/transform/load (ETL) processes. But new platforms and capabilities are rapidly becoming available and making processes more efficient and effective. And as new open-source techniques become more mainstream and commercially supported, it’s easier for businesses to take advantage of the latest innovation and data sets in a turnkey manner.
Machine learning algorithms can then be applied to mobilize and interpret the data into insights that were otherwise untapped – performing tasks like risk scoring or sentiment analysis – all while continuously optimizing insights through feedback loops.
Vodafone is a prime example of a company harnessing platform innovation in this way. To improve customer experience, we helped the firm create Intelligent Care, a solution utilizing analytics to steer customers to the best possible channel for their individual needs.
Improving the ability to predict why customers made contact enabled Vodafone to be more proactive and saved the customer from having to make the call in the first place. Less than a year after its launch, inbound calls were down by 1.5 million and digital channel use increased by 26%.
4. How do we leverage cloud services for our data platforms?
Many organizations are planning or executing an enterprise-wide “journey to cloud” strategy; these journeys tend to focus on migrating applications into the cloud to gain flexibility and reduce hosting costs. We advocate expanding this approach to focus on the incremental value from a “journey to intelligence”. This requires thinking beyond the current applications of AI and analytics within the organization to consider how to exploit new datasets in new ways.
A “journey to intelligence” approach relies on a greater influx of data to store and comes with an inherent variability in the compute processing demand – a natural fit with a shift to cloud services. However, it also raises several critical questions to be addressed in the data strategy: Which data goes to the cloud, and which doesn’t? Which cloud provider do we use? Is it better to be multi-cloud to avoid lock-in and exploit broader capabilities? How do we manage data security and movement across both on-premise and potentially multiple cloud providers?
While companies will most likely start with one cloud provider, it’s important to recognize they may end up using multiple. We’ve seen cases where companies begin to build their cloud platform with one major cloud provider – housing much of their data there – only to later realize they want to integrate capabilities from another cloud provider. As such, it’s important to design a strong multi-cloud strategy at the outset to have the right level of flexibility and modularity later should they arrive at these junctures. Without it, companies may find themselves with top-quality AI solutions that simply cannot scale efficiently.
5. Who is responsible for ethical data use?
It’s essential to have clear responsibilities for ethical data use and have a dedicated team to set the right policy, governance, and accountability structures across the entire data supply chain.
For example, when thinking about data ingestion—or the moving of data from a source to a data platform’s landing and staging area—you need to consider what data is needed and the necessary permissions. For data processing—or the modelling of data to prepare it for insight creation—you’ll need to look at what data needs to be anonymized or encrypted.
In the insight creation phase which uses analytical models, machine learning algorithms and visualizations, you should think about how to avoid bias in how the insights are generated. And with insight publication and use of insights by business users, you need to monitor that the insight is being used for the purpose intended and that the data is kept secure.
In addition to thinking about the entire supply chain, the wider organization and community need to buy into this effort: The best frameworks and policies can fail if they are not respected by everyone in the company and not communicated to consumers, some of whom may already be skeptical of AI technology. For example, if companies want to add voice recordings into their innovation platform, customers must be made aware of this and given the chance to opt out. This principle is at the core of new regulations such as GDPR.
Without ethical and responsible use, data strategies and AI solutions might work technically, but may not deliver the expected outcome.
A final thought
Having a data strategy to underpin your AI strategy is critical for competitive advantage and will ultimately help accelerate your time to value. In fact, 72% of Strategic Scalers (those who are successfully scaling AI in their organizations) said a core data foundation has been key to their success.
Answering these five questions as you embark on a new – or updated – data strategy will help ensure you’re well placed to scale with confidence and speed.