How Harvey Nash is embracing ‘Citizen AI’ for faster experimentation and adoption
Harvey Nash CIO Ankur Anand provides a case study on how the company is democratising AI
Harvey Nash CIO Ankur Anand explains how his company is using Citizen AI: an approach to enterprise AI that enables employees outside IT to create AI tools, agents and prototypes within defined guardrails.
As artificial intelligence continues to transform how businesses operate and people work, a new concept is gaining increasing traction: Citizen AI.
Put simply, Citizen AI is about putting more power into the hands of a company’s ‘citizens’ – its staff. It’s a development strategy that makes AI tools available to employees outside IT to create their own solutions, controlled by clear guardrails and parameters. This democratises AI by enabling business users to solve problems and create efficiencies in their roles without the overhead of full IT-led development.
Like citizen development in the low-code era, Citizen AI shifts some technology creation out of IT and into the business, but generative AI and AI agents lower the technical barrier significantly further.
This is an approach we have been employing at Harvey Nash for a number of months now, joining a growing cohort of businesses that are taking the same tack.
Empowering staff, encouraging experimentation
For us, Citizen AI is as much about creating a culture of innovation as it is about technology. We want to put AI into the hands of the people who understand our clients, processes and business problems best, and give them the freedom to experiment within clear guardrails.
What’s the need for this, you might ask? Historically, turning a good idea into a technology solution required specialist skills, funding and space on a development roadmap. But AI lowers that barrier. People can explore an idea, create an agent or prototype, and establish whether it genuinely solves a business problem before the organisation commits significant resources. Citizen AI taps into that ability.
The objective isn't to create hundreds of unofficial production systems, but rather to support experimentation and make it easier for good ideas to emerge from anywhere in the organisation.
For example, we want someone in Legal, Finance, Sales, Operations or another part of the business to be able to spot an opportunity and then use AI to prove that it could work and be scaled across the business or department.
Once the concept is proven, the model changes. At that point, Technology helps turn the idea into something enterprise-ready by addressing the architecture, security, data protection, integration, testing, resilience, support and ownership needed to operate at scale.
Guardrails for a safe environment
Guardrails are essential to making the model work (I’ve highlighted five below), although they shouldn't make experimentation so difficult that innovation stops. We provide an approved enterprise AI environment, controls around sensitive information and expectations around appropriate human judgement. That creates a safer space in which people can experiment without every experiment having to meet the standards of a production system from day one.
AI literacy still needed
Skills are also changing. The most important capability for an AI citizen isn't necessarily coding; it's understanding a problem well enough to describe it, knowing the context and data behind it, and having the judgement to recognise whether the result is useful.
People still need AI literacy. They need to understand data sensitivity, test and challenge AI outputs, recognise where AI can get things wrong, and know when human accountability is essential. But increasingly, natural language is lowering the technical barrier between having an idea and being able to demonstrate it.
Literacy can’t be handed off to a training course and ticked off. Anyone who uses AI owns the work that comes out of it. If an AI output ends up in a client proposal, a candidate email or a board paper, the person who sent it is accountable, not the tool.
Managers have a part to play too. They need to make sure their teams know what data can and can’t go into an AI tool, and they need to give people time to learn. Leaders set the tone here. If the message from the top is that experimentation is welcome but careless use isn’t, people know where the line is.
Citizen AI in action
So, how are we using the approach in practice at Harvey Nash? Here is a brief example:
Marketing: turning an idea into a working solution - Our content manager has used AI to create a solution around a day-to-day content marketing need. It's a good example because it shows someone can shape and prototype a solution without being a traditional developer.
Essentially, our content manager created a content agent to support our recruiters. The agent allows our recruitment consultants to find approved posts and assets, personalise to their role and tone and post them to their social channels. This supports brand amplification without requiring consultants to create new content. We can measure the number of posts created, adding a gamification element to it to encourage usage.
Citizen AI Guardrails
Then there are the guardrails. These work on several levels:
A governed platform
We use Microsoft Copilot as our approved general-purpose enterprise AI environment, rather than encouraging employees to put corporate information into public AI services. Our policy requires other AI applications used for company business to be approved by Group Technology and specifically protects company, client, candidate, supplier and personal information.
Data access is intentionally constrained
One of the most important principles is that an AI agent should not become a back door into information someone should not see. Our work on AI readiness therefore puts considerable emphasis on data permissions and ownership.
Not every employee or agent receives the same entitlement
We distinguish between general AI use, agent creation and more advanced capabilities. Access can therefore reflect what somebody is trying to achieve rather than simply switching everything on for everyone. Our internal guidance distinguishes between straightforward agents that business users can create themselves and more complex agents involving integrations that require support from the technology team.
We govern consumption as well as security
There's another guardrail that's easy to overlook: cost. Agentic AI increasingly has consumption-based economics. We therefore monitor usage rather than treating AI as unlimited computing.
IT helps an idea graduate into a responsible agent
Once something moves beyond personal experimentation, our agent requirements approach asks much more structured questions around the business problem, users, data sources, data quality and permissions. That is where IT can help the creator think through areas they may not naturally consider: permissions, sensitive information, integrations, testing, operational ownership and whether an agent is actually the right answer.
Guardrails only work if people know they exist and feel responsible for them. Every agent that moves beyond personal use needs a named owner who answers for what it does, what data it touches and whether it still does its job. Owners review access as roles change, and they switch off agents that are no longer used or no longer safe. The rules belong to the people who build with AI, not just to the people who write the policy.
A changing future
Ultimately, I see the role of IT changing. We don't need to originate every innovation or build every prototype ourselves. Our role is to create the environment in which innovation can emerge safely, help identify the proof points that matter, and then turn the strongest ideas into secure, resilient enterprise capabilities.
In summary, Citizen AI is about dramatically increasing the number of experiments an organisation can run. Give the people at the coalface the freedom and guardrails to turn ideas into proven concepts. When an idea demonstrates real value, that's when Technology steps in to make it enterprise-ready and scale it. That way, you get grassroots innovation without creating a technology wild west.
Ankur Anand is Group CIO at Harvey Nash.