Jack Barnes on how Valorem Reply is helping clients prepare for AI transformation

A subsidiary of Reply, Valorem Reply works with clients to make the most of digital transformation, and AI tools. Jack Barnes, data and analytics lead at the firm, reflects on the factors which set Valorem Reply apart from the competition – and how the future of agentic AI may unfold in the workplace.

You joined Valorem Reply six months ago. How have you adjusted to life at the firm, and how does it differ to your previous roles?

Joining Valorem Reply has been a really positive experience. There’s a strong sense of team here, and everyone has been incredibly welcoming, which made it easy to settle in quickly. In some ways, the role feels familiar because I’m still working with clients, understanding their challenges and helping them move forward. The main difference is the focus: I’ve moved from Adoption and Change Management into Data and Analytics. I worked in data earlier in my career, so part of the transition has been about refreshing that knowledge and getting back up to speed with how the technology, the market and client expectations have moved on.

At Insight, you led the senior practice team for adoption and change managment. At DXC Technology, as a digital transformation principal, you helped establish a transformation centre in the North East of England. How have your previous positions prepared you to be Valorem Reply’s data and analytics lead?

 Both roles prepared me in different but complementary ways. At DXC, my digital transformation role gave me a lot of experience working directly with clients, running workshops, shaping strategy and turning broad business challenges into practical solutions. At Insight, leading the adoption and change practice gave me the management experience that comes with running a large European consulting team; from building capability and supporting people, to managing demand when there is either too much work or not enough. That mix of client-facing transformation experience and practice leadership has been really valuable in helping me step into the Data and Analytics lead role at Valorem Reply.

How does your work interact with the wider Valorem Reply team? And how does the wider Reply group support what you do?

Data and analytics touches almost every part of what Valorem Reply does, because it sits at the centre of so many client conversations. Within the UK team, I work closely with colleagues across platform, integration, security, AI and applications, because good data outcomes rarely sit in isolation. A client might come to us wanting better reporting or a Microsoft Fabric roadmap, but very quickly that connects into their Azure foundation, integration landscape, governance model, security posture and wider AI ambitions. That is where the broader Valorem Reply team is so valuable: we can bring the right specialists together around the client problem, rather than treating data as a standalone workstream.

The wider Reply group gives me access to a much bigger network of expertise, experience and credibility. Reply is a decentralised group of specialist companies, so there is always someone who has seen a similar challenge before, whether that is in data strategy, Microsoft Fabric, AI, cloud platforms, integration or industry-specific transformation. That means we can combine the agility of a focused UK team with the depth of a much larger organisation. For me, that support is a real advantage: it helps us bring proven thinking, case studies, accelerators and specialist knowledge into client conversations, while still keeping the delivery personal and practical.

The AI era has supposedly been upon us since late 2022 – and yet almost four years into the hype, many firms still seem unclear one what it should look like for them. Where are the areas they should focus first, when it comes to actually seeing returns on investment?

I think the starting point has to be value, not technology. The organisations seeing real returns from AI are usually the ones that start with a clear business problem, not a broad AI ambition. That could mean reducing manual effort in reporting, helping employees find knowledge faster, improving service responses, or supporting better decision-making. The important thing is that the outcome can be measured in terms of time, cost, quality or risk.

For most clients, the best first step is to focus on practical areas where AI can augment existing work, such as knowledge management, automated insight generation, reporting and analytics, customer or employee service, and developer or analyst productivity. At the same time, they need to invest in the data foundation underneath it. If the data is fragmented, poorly governed or not trusted, AI will only expose those issues faster. So the firms making progress tend to pair targeted, near-term AI use cases with a governed data platform that allows them to scale safely.

When it comes to data and AI foundation, what does an AI-ready data platform look like?

An AI-ready data platform is not just a place to store data. It is a governed, trusted foundation that makes data usable for analytics, automation and AI in a safe and repeatable way. At the simplest level, it means bringing data together from the right sources, organising it properly, applying clear ownership and governance, and making sure people can understand where the data has come from, how it has changed and whether it can be trusted.

In practice, that means a platform with strong foundations that include:

• Secure access.
• Good data quality.
• Clear lineage.
• Reusable data products.
• Well-designed semantic models.
• Sensible controls around sensitive information.

It also needs to support both traditional reporting and newer AI use cases, so the same governed data can be used for dashboards, Copilot-style experiences, machine learning models or agentic workflows. If every AI use case needs its own bespoke data preparation, it becomes slow, expensive and hard to control.

For me, the test of an AI-ready platform is whether the business can confidently use its data to make decisions and automate work without constantly questioning the source, quality or security of that data. The technology matters, of course, but the real differentiator is trust. If the platform gives people trusted data, governed access and a clear route from idea to production, then AI has a much better chance of delivering real value rather than becoming another isolated experiment.

A further part of this is agentic AI. What are the key differences when people refer to generative and agentic AI, and what practical results have you seen from clients putting agentic AI to work?

The simplest way I explain it is that generative AI helps you create or summarise something, whereas agentic AI helps you get something done. Generative AI might draft a response, summarise a document or answer a question based on available information. Agentic AI goes a step further because it can reason through a task, decide what actions are needed, use tools or data sources, and move a process forward within the boundaries you set for it.

In practical terms, the strongest use cases I have seen are where agents take the friction out of repeatable knowledge work. That could be helping a service team triage requests, supporting analysts by pulling together information from different systems, guiding employees through internal processes, or helping commercial teams prepare for client conversations. The value is not that the agent replaces the person; it removes some of the manual searching, checking and coordination that slows people down.

The best results come when clients start with a contained process, clear guardrails and a visible measure of success. For example, reducing the time it takes to find the right information, improving the consistency of a response, speeding up a workflow or freeing specialists to focus on higher-value work. Agentic AI becomes powerful when it is connected to trusted data and business systems, but it still needs good governance, human oversight and a clear understanding of where automation is appropriate. Done well, it can make teams more effective very quickly; done without those foundations, it risks becoming another interesting pilot that never really scales.

What advice would you give to a company who has yet to take the leap into Agentic AI? Where is the best place to start – and what are the pitfalls to beware?

My advice would be to start small, but start properly. Agentic AI works best when it is applied to a well-understood process where there is enough repetition to create value, but enough human judgement involved that people still need support. I would not begin with the most complex, business-critical workflow in the organisation. I would start with something contained, such as internal knowledge search, service request triage, report preparation, sales or bid support, or helping teams navigate policy and process information more easily. The key is to choose a use case where the outcome is visible and measurable, whether that means reducing the time it takes to find information, improving the consistency of responses, speeding up a workflow or removing manual hand-offs.

The main pitfalls are usually less about the technology and more about discipline. Organisations can move too quickly into automation before they understand the process, connect agents to poor-quality or poorly governed data, or underestimate the importance of security, access control and human oversight. There is also a danger of treating agentic AI as a replacement for people rather than a way to help them work more effectively. The best place to start is with a focused problem, trusted data, clear guardrails and a human-in-the-loop model. That gives the organisation confidence, proves value quickly and creates a pattern that can be reused safely elsewhere.

Reply’s members have an excellent relationship with Microsoft. What are the benefits which Fabric’s unified analytics platform brings, and how do you help your clients to make the most of them?

The big benefit of Microsoft Fabric is that it brings a lot of the analytics landscape together into one platform. For many organisations, data is still spread across different tools, teams and processes, which makes it harder to govern, harder to trust and harder to turn into value. Fabric helps by giving clients a more unified way to manage data engineering, data warehousing, real-time analytics, Power BI, data science and AI use cases, all sitting on top of OneLake. That matters because it reduces some of the friction between teams and creates a clearer path from raw data through to insight, automation and AI.

The way we help clients make the most of it is by not treating Fabric as just a technology deployment. We start with what the organisation is trying to achieve: the decisions they need to improve, the reports they rely on, the data sources causing pain, and the AI or Copilot use cases they want to unlock. From there, we help them shape the right architecture, governance model and delivery roadmap. That could mean assessing Fabric readiness, designing the lakehouse and warehouse patterns, building governed data products, improving semantic models, or helping teams adopt better ways of working around data ownership, quality and security.

Our relationship with Microsoft is important because it means we are close to where the platform is going, but the value for clients comes from turning that knowledge into something practical. Fabric can simplify the data estate, accelerate reporting, create a governed foundation for AI and reduce duplication across teams, but those benefits only come through good design, strong governance and real adoption. Our role is to help clients move from interest in the platform to measurable outcomes, so Fabric becomes a foundation for better decisions and future AI capability, rather than another tool in an already crowded data landscape.

Finally, when it comes to implementing change on any of these fronts, you must encounter pushback. With mounting pressure on new technologies to demonstrate meaningful improvements at work, how does Valorem Reply help clients to respond to criticism from the boardroom or the wider office, during change implementation?

Pushback is normal during data change, because it often affects how people report, make decisions and trust information they have used for years. We help clients respond by making the change practical and evidence-led. Rather than talking about “data transformation” in broad terms, we focus on the specific outcomes: better quality reporting, clearer ownership, faster access to trusted data, reduced manual effort and stronger governance.

For the boardroom, that gives a clearer way to judge value. For the wider organisation, it helps people understand why the change matters to their day-to-day work. We involve users early, show what will improve, and use focused use cases to prove the benefit before scaling. That way, criticism can be answered with real evidence, and data change becomes less about new technology and more about helping people work with better, more trusted information.

In many ways, data changes are similar to AI transformation more broadly: the technology only succeeds when people trust it, understand it and can see how it improves the work they already do. That is why adoption, governance and measurable value matter just as much as the platform itself.