Key takeaways_

• Start with built-in AI agents in your existing platforms before investing in custom solutions to unlock quick value with minimal effort.
• Unified, secure data and embedded governance are essential for effective, trustworthy AI agent deployment.
• Empower your people to experiment and iterate, measuring results to drive ongoing improvement and scalable impact.

AI agents are everywhere in the headlines, promising to revolutionise the way we work. Yet for most organisations, the reality is more confusing than transformative. The hype is enormous, but practical guidance on how to make AI agents deliver real value is often missing. Many teams are left asking: Where do we start? Do we need to build everything from scratch? How do we avoid risk?

At Infinity Group, we faced these same questions. Rather than getting lost in the noise, we focused on what would actually work for our business. By leveraging our Microsoft tech stack and a pragmatic approach to AI adoption, we’ve made AI agents a genuine driver of growth.

The results speak for themselves. Six months after replatforming to an all-Microsoft tech stack, we achieved 35% growth without increasing headcount.

We share how we made AI agents work for us, and what your organisation can learn.

What are AI agents – and why all the hype?

AI agents are digital assistants that perform tasks, make decisions or automate processes. They often work alongside humans to boost productivity and efficiency.

In the business world, they can qualify sales leads, automate invoice processing, summarise meetings or even help resolve customer service issues.

The hype around AI agents is driven by their promise to transform how organisations operate:

• Efficiency gains: Automating repetitive tasks frees up people for higher-value work.
• Smarter decisions: AI agents can surface insights from vast amounts of data in real time.
• Scalability: Businesses can grow without adding headcount, as AI handles more of the workload.

Despite the promise, many organisations stumble at the first hurdle. Common pitfalls include:

• Chasing the latest trend: Jumping into AI without a clear business case or understanding of what’s already available in existing platforms.
• Overcomplicating the approach: Believing everything must be built from scratch, rather than leveraging powerful out-of-the-box agents.
• Neglecting data and security: Deploying agents without unified, well-governed data can lead to poor results and compliance risks.
• Forgetting the people: Failing to empower staff to experiment and adopt AI in their daily work means missed opportunities for real impact.

The key is to start with what you have, focus on real business needs and build a foundation that’s secure, integrated and ready to scale.

Our step-by-step blueprint_

Over the last year, we’ve experimented with AI agents, creating the right foundations across our processes and data and seeking meaningful use cases that add genuine value. Here’s what we’ve learned in that time to accelerate your journey.

Step 1: Utilise first-party agents_

Many organisations jump straight into building custom AI solutions. But this often leads to wasted time, duplicated effort and missed opportunities. So, our journey began by making the most of our existing technology.

Because Infinity Group is fully platformed on Dynamics 365, we had immediate access to a suite of out-of-the-box AI features – such as Sales Qualification Agent, Payables Agent and Time Entry and Approval Agent.

Before considering any custom development, we explored and enabled these built-in tools. This approach not only accelerated value but also ensured we were working with solutions already designed to integrate with our core data sources.

This gave us an easy opportunity to test out AI agents across different areas of the business, without having to invest time and money into complex dev work.

Step 2: Prioritise data integration and security_

AI agents are only as good as the data they use. Siloed, inconsistent or poorly governed data leads to unreliable results and increased risk. A unified, secure data foundation is essential for effective, trustworthy AI.

We treated preparation as a discrete phase before heavy customisation:

• Define sources of truth and lineage: We catalogued key datasets (e.g. customer, opportunity, ticketing, finance) and their authoritative systems, then documented how data flows into Dataverse and OneLake so agents can trust what they read.
• Profile and cleanse data: We standardised formats, removed duplicates and enforced naming conventions so prompts and retrieval logic return consistent results.
• Model the process, not just the tool: For each target workflow (e.g. pre‑mortem risk checks, SDM service updates, payables), we drew swim‑lane maps, decision points and exception paths. That gave agents a map to follow rather than a vague task description.
• Plan governance upfront: We created a simple taxonomy and sensitivity model in Purview (labels, policies, retention) and aligned Dynamics 365 security roles/field‑level security to keep access appropriate for the agent’s scope.
• Set quality KPIs: We measured data completeness, consistency and timeliness per domain, so teams can see when data quality – not the agent – is limiting outcomes.

This upfront preparation meant our agents could reason over the right context (documents, transcripts, records) and execute processes reliably from day one, instead of learning hard lessons in production. It also created a repeatable template for future agents.

Step 3: Customise and extend where needed_

Every business has unique needs, and sometimes standard solutions need to be tailored.

We reviewed where first-party agents fell short and identified opportunities for customisation. For example, we connected Customer Service Copilot to Azure DevOps and Microsoft’s support documentation, so our service desk engineers could reason over a customer’s DevOps backlog and apply best practice guidance from Microsoft Learn – all within Dynamics 365 Customer Service.

When a business need wasn’t met by existing tools, we built new solutions using low-code platforms like Copilot Studio and Power Automate. A standout example is our Risk Identification Agent, which analyses call transcripts to flag potential project risks during sales and presales cycles. This agent uses Microsoft Graph, Power Automate and AI Builder to extract and store risks directly against opportunity records in Dynamics 365.

Step 4: Prioritise data integration and security_

With preparation complete, integration and governance become ongoing disciplines. This ensures that AI agents have access to a single, unified view of the business – eliminating the confusion and inefficiency that comes from fragmented or duplicated information. Security and governance are also non-negotiable in a world of increasing regulation and cyber risk.

To overcome these challenges, we unified our business data into OneLake, Microsoft’s enterprise data lake, which acts as a single source of truth for all departments. This integration broke down silos and enabled real-time insights, so AI agents could reason over the most current and complete information available.

We embedded governance from the outset using Microsoft Purview, which allowed us to classify, secure and audit data by design. This included setting up sensitivity labels, access controls, and retention policies to ensure that only the right people (and the right agents) could access specific information.

Regular tenant health and cost checks, using tools like Surveil and Digital Insights, became part of our operational rhythm. These checks help us spot anomalies, optimise resource usage and ensure that our AI agents are always operating on reliable, secure, and up-to-date data.

Step 5: Empower people to experiment_

When it comes to AI, people lead the charge. Humans are crucial in finding use cases for agents, putting them into place and verifying the value on their daily work. To make experimentation with AI agents a reality, we set out clear policies and launched targeted initiatives designed to empower every employee, regardless of technical background.

To bring it to life, we launched several initiatives:

• Personal AI audits: Employees were encouraged to review their own roles and workflows to identify where AI agents could save time, automate repetitive tasks, or unlock new insights. These audits helped staff see practical opportunities for AI in their day-to-day work, driving grassroots adoption and enthusiasm.
• Training and support programmes: We rolled out structured training sessions and digital resources so everyone could learn how to use, customise and get the most from AI agents. This ensured that technical expertise wasn’t a barrier to participation.
• Dragon’s Den innovation pathway: For custom solutions, we set up a “Dragon’s Den” process where teams could pitch ideas for new agents. Approved ideas were prioritised and developed by our internal delivery team, giving staff a clear route from concept to implementation and ensuring that innovation was both bottom-up and strategically aligned.
• Cross-team sharing: Wins and lessons are actively shared across departments, breaking down silos and accelerating adoption. We celebrate successful experiments and openly discussed failures, creating a safe space for learning and iteration.

These policies and initiatives created an environment where experimentation was actively encouraged.

Step 6: Measure and iterate – and scale_

Without measurement and iteration, it’s impossible to know what’s working or where to improve. Continuous feedback ensures AI investments deliver ongoing value and adapt to changing needs.

We track key metrics such as time saved, business growth, project profitability and customer satisfaction scores. And this is complemented by qualitative feedback from users, ensuring we capture both the hard numbers and the lived experience of staff and clients.

This is an ongoing exercise as we delve deeper into AI agents, ensuring that efforts are warranted and rewarded with genuine business value. Once we have that, it becomes simpler to scale usage and find similar use cases to increase results.

By making measurement, iteration and scaling an ongoing discipline, we ensure that our AI strategy remained agile, impactful and futureproof.

The AI agents we use – and the results_

At Infinity Group, our approach to AI agents is both pragmatic and ambitious: we leverage a blend of powerful first-party solutions, smart customisations and innovative in-house developments to drive measurable business value.

We began by enabling a suite of first-party agents within Dynamics 365, each designed to automate and enhance core business processes:

• Sales Qualification Agent and Sales/Opportunity Research Agents streamline the sales cycle, qualifying leads and surfacing insights so our teams can focus on the most promising opportunities.
• Payables Agent automates invoice processing and approvals, reducing manual workload and accelerating cash flow.
• Time Entry and Approval Agent ensures accurate, timely recording of work, supporting both compliance and resource management.
• Knowledge Management Agent and Case Management Agent help us capture, organise and resolve customer and internal queries efficiently.
• Customer Service Copilot and Sales Copilot provide real-time support and recommendations, empowering our teams to deliver exceptional service and close deals faster.

Custom agents_

Recognising that every organisation has unique needs, we also customised and extended these agents:

• Our Customer Service Copilot is integrated with Azure DevOps and Microsoft Learn, enabling service desk engineers to reason over a customer’s DevOps backlog and apply best practice guidance – all within Dynamics 365 Customer Service. This integration bridges the gap between technical delivery and customer support, ensuring issues are resolved quickly and knowledgeably.
• The Risk Identification Agent is a standout example of our innovation. By analysing call transcripts from sales and presales conversations, this agent detects nuanced signals of potential delivery risks (such as missing project management or data challenges) and stores these findings directly in Dynamics 365. This proactive approach helps us address issues before they escalate.
• The Minute Master Virtual SDM Agent synthesises information from call recordings, documents and Dynamics 365 data to generate comprehensive service update reports for customers, saving time and ensuring consistency.
• Our Talent Development Agent, built with Copilot Studio, supports recruitment by assessing cultural and skills fit during candidate screening, helping us build stronger teams.
• The Product Knowledge Agent automates responses to internal queries, drawing on technical documentation and internal resources to provide instant, accurate answers – freeing up expert time and improving self-service across the business.

Results delivered_

The impact of these AI agents has been transformative:

• 35% business growth achieved without increasing headcount, demonstrating true scalability.
• 180+ hours saved weekly across the business, as automation and AI take on repetitive tasks.
• 16% rise in customer satisfaction scores, reflecting faster, more personalised service.
• 15 hours saved on billing processes alone, reducing admin overhead and accelerating cash flow.

These results are not just numbers: they represent a fundamental shift in how Infinity Group operates. By combining out-of-the-box capability with targeted customisation and a culture of continuous improvement, we’ve built an AI-powered foundation for sustainable growth, resilience, and exceptional customer experiences.

Our top tips for other organisations_

Drawing on our journey with AI agents, here are our most valuable lessons for organisations looking to achieve real impact, rather than aimlessly follow the hype:

1. Start with what you have_

Most modern platforms now include a wealth of powerful AI agents out of the box. Microsoft is regularly putting out pre-built agents across their products, such as Dynamics 365.

Before investing time and resources in custom development, enable and experiment with these built-in tools. You may be surprised by how much value you can unlock simply by activating and configuring what’s already available. Plus, it means minimal effort and no extra cost.

2. Customise, don’t reinvent_

While first-party agents are a great starting point, every organisation has unique needs.

Tweak and extend existing agents to fit your workflows and business context, and see how that impacts the output you gain.

Only invest in building bespoke solutions when there’s a clear, strategic business case, as this keeps your AI programme focused and cost-effective.

3. Integrate data and embed security_

Unified, secure data is the foundation for effective AI. Prioritise breaking down silos and ensuring your agents have access to accurate, up-to-date information. Ideally, you should do this before you implement AI agents if you want the best possible output.

Embed governance and security from the outset, using tools like Purview, so compliance and risk management are built in – not bolted on as an afterthought. Establish clear policies, roles and controls for how data is used and how agents operate. This not only protects your organisation but also builds trust with customers and stakeholders. Our guide to data governance in the age of AI can help you achieve this.

4. Empower experimentation_

AI adoption thrives when people are encouraged to try new things.

Give your teams the freedom, training and support to experiment with agents in their own workflows. Celebrate both successes and lessons learned, and make it easy for staff to share their experiences across the organisation.

5. Use a structured process for custom ideas_

Innovation needs direction. At Infinity Group, we use a ‘Dragon’s Den’ style board to evaluate and prioritise custom agent ideas. This ensures only the most valuable and feasible concepts are developed and scaled, while giving everyone a clear route from idea to implementation.

6. Measure and iterate_

Track the results of your AI initiatives – whether it is time saved, customer satisfaction or business growth.

Use these insights to refine your approach, scale what works, and quickly pivot away from what doesn’t. Continuous improvement is key to long-term success.

7. Share success stories_

Nothing accelerates adoption like real-world wins.

Publicise your successes, highlight the teams and individuals driving change, and make lessons learned accessible to all. This helps build momentum and encourages others to get involved.

Build AI agents for resilience and scalability_

Transformation isn’t a one-off project. It’s a mindset and an ongoing journey. At Infinity Group, we’ve learned that true, sustainable growth comes from integrating technology, intelligence and enablement across every part of the business.

By focusing on unified data, robust governance and a culture of experimentation, we’ve built a foundation that not only delivers results today but is ready to adapt and scale for whatever comes next.

The lesson is clear: success with AI agents isn’t about chasing the latest trend or building everything from scratch. It’s about starting with what you have, empowering your people and creating a connected, intelligent platform that supports continuous improvement and innovation.

Ready to make AI agents work for you?

For practical advice, real-world examples and a step-by-step roadmap, download our Ultimate Guide to AI Agents – the essential resource for organisations looking to unlock the full potential of AI.