Computer Weekly’s June 20, 2018 interview with Mike Young, then CIO of Centrica, described a transformation strategy built on three linked ideas: create a usable data foundation, apply it to customer and operational problems, and change the organisation so employees can act on the results.
The interview covered Centrica’s data lake, machine-learning projects, Microsoft and SAP partnerships, ERP modernisation, field-service systems and digital adoption. It is a historical account of plans and reported results at that time—not evidence of Centrica’s technology estate, leadership, vendor relationships or transformation outcomes in 2026.
As an Amazon Associate I earn from qualifying purchases.
The strategy in one sentence
Young’s approach was not simply a cloud migration or an artificial-intelligence programme. It combined platform modernisation, customer-service improvement, potential data-services revenue and organisational change.
The framework was particularly clear in Young’s description of transformation as a three-part process:
#1 Best Overall
- Wiley
- Language: english
- Book - storytelling with data: a data visualization guide for business professionals
- Use the right technology to produce the required data.
- Wrap that technology in an effective business process.
- Equip and train the people who must use it.
That distinction matters. A data lake does not improve a customer journey by itself, and a faster ERP does not create value if front-line teams cannot use the information it produces.
Why Centrica needed a different technology model
In the interview, Centrica was described as a large and complex group with more than 23 million customers, eight significant UK ERP systems and two major SAP platforms. Young said some back-end processes were constrained by batch processing that could leave data up to 24 hours behind real time.
The company also reportedly had about 50,000 field engineers using 11 different systems. That combination created several distinct problems:
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →- Customer and operational data was distributed across legacy platforms.
- Back-end information was not always available quickly enough for modern digital service.
- Field workers had to navigate fragmented tools.
- Digital channels, contact centres, billing and field operations were not necessarily working from the same current information.
These conditions explain why the programme had several strands. It was simultaneously an efficiency effort, a customer-service programme, a platform-modernisation initiative, a possible data-services business and an attempt to build more software-driven operating capabilities.
The source does not establish that every initiative applied uniformly across Centrica or its British Gas operations. It describes the strategy as Young presented it in 2018.
From a data lake to operational capability
Young said Centrica’s data journey began at the end of 2014 and that a data lake was in place by 2015. The environment included technologies such as MapReduce and Spark and brought structured and unstructured data together.
The stated objective was more ambitious than creating a large repository. Centrica’s teams used the resulting information to develop proprietary algorithms and identify business applications. Young described a data-science organisation of about 300 people, including analytics and visualisation specialists.
Recommended Free Tools
A notable feature was the mix of skills. The group included highly qualified data specialists as well as people with field-engineering and customer-operations experience. That blend is important because many utility problems are not purely analytical. A technically elegant model may fail if it does not reflect how engineers work, why customers call or where a process breaks in practice.
The customer-journey machine-learning example
The interview’s clearest operational example involved a machine-learning system that tracked customer activity online. It showed how far customers progressed through a digital journey, identified where journeys failed and helped contact-centre and operations teams understand why customers subsequently called.
Young said an initial release took approximately six weeks. The system was then refined for contact-centre deployment and was reported to have reduced repeat failures. The article did not provide a baseline, sample size, independent validation or quantified improvement, so that outcome should be treated as Young’s account rather than a verified benchmark.
The significance of the example is not the label “machine learning”. It is the feedback loop:
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problems- Observe what customers attempt online.
- Identify where the journey stops working.
- Connect that failure to contact-centre demand.
- Give operations teams information they can use to correct the process.
This is a practical model for data science: start with a measurable service problem, deliver a useful minimum viable product quickly and refine it with operational users.
Io-Tahoe and the data-services experiment
Centrica also explored whether its internal data capabilities could become a commercial product. The interview said the company had established a £100m venture-capital arm, acquired data-tools company Rokitt-Astra and combined that capability with its own data work in a spin-off called Io-Tahoe.
The proposed market included banks and organisations facing data-discovery and governance challenges associated with GDPR. This should not be read as evidence that Centrica intended to sell customer data indiscriminately. The reported proposition was data services and data discovery.
GDPR did not create a guaranteed market. Rather, requirements around finding, understanding and governing personal data created a potential business opportunity for tools that could help organisations identify what data they held and where it was located.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →The 2018 interview described the strategy and intended opportunity. It did not establish Io-Tahoe’s long-term commercial success, later ownership, revenue or continuing relationship with Centrica.
Microsoft: strategic access rather than proof of technical superiority
The interview described a strategic relationship between Centrica and Microsoft. Centrica’s chief executive Ian Conn had met Microsoft CEO Satya Nadella, and Centrica was participating in a strategic utilities initiative. Young described Centrica as Microsoft’s only energy-sector partner among three selected key partners at the time.
Centrica’s chief data scientist also sat on a Microsoft advisory team. Young said the relationship could provide access to pilots, cloud expertise and Microsoft’s data-science capability, including access to what he described as 300 data scientists in Redmond.
Those statements describe the expected value of a 2018 partnership. They do not prove that Azure was technically superior to alternatives, that the arrangement remained active in the same form or that Microsoft supplied a contractual team of 300 people to Centrica.
Strategically, the partnership represented an attempt to reduce the time and risk involved in experimenting with artificial intelligence, machine learning and cloud services. The trade-off was greater dependence on a major platform vendor and its roadmap.
SAP HANA and the problem of legacy ERP
Young said Centrica had already moved its central financial ERP from legacy SAP to SAP HANA. He also described a broader plan to move legacy systems toward HANA over approximately two years.
Rank #4
According to the interview, the financial system could update data every second, while older environments had been affected by batch-processing delays. Young reported that a processing path had been reduced from a 24-hour limit to approximately one hour.
That one-hour figure needs careful interpretation. The source does not establish that every Centrica system, data class or business process operated at that latency. It is a reported outcome for the context Young was discussing, not a universal performance guarantee.
The wider architecture also included SAP Ariba for procurement, Workday for human resources, an API layer and machine-learning processes intended to help inspect and repair back-end data. The interview separates what had reportedly been completed from what was still planned:
| Area | Status described in 2018 |
|---|---|
| Central financial ERP | Migration from legacy SAP to SAP HANA was reported as complete. |
| Broader legacy estate | A wider migration was planned over approximately two years. |
| Back-end billing and ERP work | Described as difficult and ongoing. |
| Processing speed | A reduction from a possible 24-hour batch delay to about one hour was reported by Young. |
Moving to HANA could improve speed and data availability, but it would not remove the harder work of data cleansing, integration, testing, process redesign and change management.
Field-service technology: the practical test
Centrica’s field-service estate showed why central data modernisation was only part of the transformation. About 50,000 engineers were reportedly using 11 different systems, and Young said the company planned to introduce a new cloud-based field-management system over the following two years.
A system of that kind would typically need to coordinate scheduling and dispatch, engineer availability and skills, customer appointments, mobile connectivity, work orders, asset histories, parts, billing and legacy integrations. Those are analytical implications of the reported problem, not features confirmed by the interview.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
The case illustrates the difference between three layers of transformation:
- Data modernisation: making information more accessible and current.
- Customer-channel improvement: reducing failed journeys and unnecessary contact.
- Front-line enablement: giving engineers and operations teams usable tools.
Replacing fragmented field systems can be more difficult than launching a pilot because it affects scheduling, safety, workforce routines, customer promises and back-end financial processes at the same time.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.The people problem was harder than the technology
Young identified adoption as the most difficult part of transformation. Centrica’s reported maturity model had three levels:
- Introductory lunchtime sessions explaining what “digital” meant for the company.
- Digital initiatives designed to help employees perform their jobs.
- A digital leadership group that could be deployed across the business to investigate transformation opportunities.
The company was also reported to have approximately 600 digital champions. That model suggests a combination of central coordination and distributed adoption: a core group develops capability, while champions help translate it into business units.
Young offered a rough adoption rule of thumb in which one-third of employees adapt quickly and become digital agents, one-third need coaching and time, and one-third never fully adapt. This is a management heuristic attributed to Young, not a validated universal ratio. Adoption depends on role, incentives, leadership, training quality and whether the new system actually makes work easier.
A champion network can be valuable, but only if its members have time, authority, training and access to delivery teams. Otherwise, “digital champion” risks becoming a symbolic title rather than an operating mechanism.
Office 365 as the collaboration layer
In the 2018 environment described by Young, Centrica used Microsoft Office 365 as a collaboration stack that included email, Skype for Business, Microsoft Project and Power BI.
These product names should be understood historically. The interview was describing the tools and branding in use at that time, not current Microsoft product naming or Centrica’s present collaboration environment.
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchPC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11What the strategy got right
- It connected platforms to business problems. The customer-journey example tied data science to failed digital transactions and contact-centre demand.
- It valued speed to learning. A six-week initial release allowed the organisation to learn before attempting a larger deployment.
- It combined specialist and domain knowledge. Field and customer-operations experience can be essential when models must change real work.
- It treated adoption as part of transformation. Training, champions and maturity-building were not presented as afterthoughts.
- It linked architecture with commercial experimentation. Io-Tahoe represented an attempt to turn internal data capability into a potential external service.
What remained unproven
The interview was strong on executive strategy and named technologies but limited on independent evidence. It did not establish:
- Whether the planned SAP HANA migration was completed.
- Whether the cloud field-management programme met its timeline or improved engineer productivity.
- Whether Io-Tahoe became a durable and successful business.
- Whether the Microsoft relationship continued in the same form.
- How the reported one-hour processing result was measured or how widely it applied.
- The baseline, sample size or independent validation behind the repeat-failure claim.
- The total transformation cost, expected return or payback period.
- The operating model for architecture governance, funding, product ownership and accountability.
There were also important issues that received little attention in the interview: cybersecurity, resilience, privacy, data lineage, access control, retention, consent and the separation between customer-service data and any external commercial use.
The CIO lessons from the 2018 account
- Start with an operational failure, not an AI label. Define the customer or employee problem, the baseline and the action the new information will enable.
- Build data quality into the operating model. A lake or faster database is not the same as trusted, governed data.
- Use pilots to learn, then industrialise deliberately. Rapid delivery is valuable, but production systems need security, ownership, monitoring, resilience and maintainable integrations.
- Match processing speed to the decision. Not every use case needs real-time data; where it does, the requirement must be designed across ERP, APIs, billing and operations.
- Pair central platforms with domain expertise. Data scientists and field or customer specialists answer different parts of the same problem.
- Design adoption before deployment. Training, incentives, champions and usable workflows determine whether a system changes outcomes.
- Separate vendor access from vendor dependence. Strategic partnerships can accelerate experimentation, but they also create concentration and switching risks.
- Attribute claims precisely. A reported benefit, a planned migration and a completed capability are different categories of evidence.
Mike Young’s 2018 Centrica interview is most useful when read as a case study in joining those disciplines. Its central lesson is not that one vendor, database or machine-learning model solved transformation. It is that technology had to be connected to data quality, business process and human behaviour before it could improve customer service or operations.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Free tools Windows power users keep installed
One-click scans. No signup required.
