I Wasn’t in the Room. My AI Avatar Was. But That Wasn’t the Biggest AI Lesson.

I Wasn’t in the Room. My AI Avatar Was. But That Wasn’t the Biggest AI Lesson.

A last-minute family issue meant I could not join Amanda in person for our presentation at Surrey Property Exchange in Guildford.

So I apologised to the audience through an AI avatar and told them something entirely true: Amanda is a far better presenter than I am, and they were in safe hands.

The avatar then explained why our property business moved from basic call-answering technology to a team of specialist AI agents.

The AI avatar that stood in for me in Guildford.

That was memorable. It was also a useful demonstration of the technology.

But it was not the most important AI lesson to come out of the evening.

The bigger lesson was revealed by the audience’s questions.

Many people in the room were already using AI. Some were dabbling. Others were building their own collections of tools and agents. There was enthusiasm, confidence and a genuine appetite to understand what comes next.

Yet beneath that enthusiasm sat a more important uncertainty:

How do you turn access to powerful AI into a dependable way of running a business?

That is the gap I believe property businesses, and many other SMEs, now need to address.

The Industry Does Not Lack AI Tools

The audience wanted to know which large language model we used. They asked how to build their own agent, where to begin and how long development would take.

Those are understandable questions. Tools such as ChatGPT, Claude, Copilot and Manus have made advanced capability available to almost anyone.

But access to capability is not the same as having an operating system for your business.

A subscription can help you draft an email, summarise a document, analyse information or create an impressive prototype. It does not automatically know:

•       which procedure your team is meant to follow;

•       what information must be collected before action is taken;

•       who has authority to approve cost or risk;

•       when a tenant issue should be passed to a person;

•       where a reliable record should be kept;

•       what happens if information is missing or contradictory;

•       how you measure whether the change has actually improved the business.

That is why I would not begin with the question, “Which AI should I buy?”

I would begin with:

“Which part of this business repeatedly creates wasted time, missed work or dependence on one person?”

We Did Not Set Out to Build Seven Agents

Our journey began in a much simpler place.

We first used a basic AI call-answering service. We then explored outbound calling. From there, we developed a more capable communications agent.

The motivation was never to collect impressive technology.

We were trying to fill the gaps that appear during a busy week.

The tasks that people forget because the next urgent issue arrives.

The jobs that are performed inconsistently because somebody is stretched.

The follow-up that good people do not enjoy and therefore postpone.

The knowledge that sits inside one experienced colleague’s head.

The coaching conversation that never quite happens because there is always something more urgent.

As regulation and operational complexity increased, the scope expanded. We developed specialist agents for communication, outbound engagement, compliance, data analysis, accounting, company knowledge and management coaching.

The seven agents were not seven technology experiments.

They were seven responses to recurring business gaps.

Amanda presenting our team of specialist AI agents.

The Most Important Thing We Did Was Switch Something Off

One of the stories we shared was that we built an agent capable of completing a process from beginning to end, and then deliberately reduced its autonomy.

The technology was ready.

The business was not.

That statement can sound like a failure. I see it as one of our most valuable results.

The agent could generate more activity than the team was ready to supervise. The issue was not whether the technology could act. The issue was whether the organisation had the procedures, capacity, oversight and confidence to control what happened next.

That is the part of AI adoption that glossy demonstrations often ignore.

A business can automate an activity before it is ready to manage the consequences.

The more capable the system becomes, the more important clear responsibility becomes.

Who approves expenditure? Who deals with a sensitive complaint? Who checks an unusual compliance case? Who owns the final decision? What gets recorded? What happens when the system is uncertain?

Good implementation does not remove people from the picture. It gives people clearer points of control.

“I Can Build This Myself”

A number of people in the room were confident that they could build their own AI stack.

Some probably can.

The technology is increasingly accessible, and I would never discourage responsible experimentation. Building small tools can teach you a great deal about your own business.

However, there is an important difference between creating an agent that works in a demonstration and operating one reliably with real customers, staff and personal information.

The second challenge requires more than a clever prompt.

It requires procedures, permissions, testing, data controls, exception handling, monitoring, maintenance and an honest view of who remains responsible.

This is why our advice is not automatically, “Let us build you an agent.”

Sometimes the right answer is to improve the existing procedure.

Sometimes a ready-made product already solves the problem.

Sometimes one focused agent is enough.

Sometimes a larger property business needs several specialist agents working across clearly defined boundaries.

The commercial opportunity for Essential AI is not to force every organisation towards the most complicated answer. It is to help identify the smallest answer that creates useful, measurable change.

The Room’s Biggest Concern Was Not Maintenance

The feedback from Guildford was instructive.

The Surrey Property Exchange audience in Guildford.

Compliance, certificates and general organisation caused greater concern than maintenance. During an exercise, people were asked to picture one rental property and identify where all the current safety records were held.

Many could not answer immediately.

That does not mean they were careless landlords. It means their information was fragmented between inboxes, folders, agents, phones and memory.

This is exactly the kind of problem that tempts people to jump straight to AI.

But the first task is not to ask a model to interpret every document.

The first task is to decide what a complete property record should contain, where it should live, who maintains it and what must happen when something is missing.

AI can then help organise, retrieve, remind and surface information within those rules.

Without the rules, technology can make confusion move faster.

Data Questions Are a Sign of Progress

The audience also raised questions about tenant data, consent and what information is shared with AI services.

I was pleased they did.

Property businesses hold personal information and deal with situations that may involve vulnerability, health, finance, complaints or disputes. Data protection cannot be treated as a paragraph added after the product is built.

Nor should “AI” be used as a vague label. Businesses need to understand what information is being processed, for what purpose, through which service, with what controls and under whose responsibility.

These questions are not a barrier to innovation.

They are part of professional implementation.

As adoption grows, I expect the strongest businesses will not necessarily be those that use the greatest amount of AI. They will be those that know where it belongs, and where it does not.

Voice Made the Future Feel Normal

Landlord PA generated particular interest when Amanda demonstrated the idea of speaking naturally to a personal Property PA.

Voice changes the way people perceive a system.

A landlord may not want to navigate another dashboard. But asking, “What needs my attention this week?” or instructing the PA to follow something up feels familiar.

Landlord PA’s functionality is live across voice, email and calls. The strategic value is not simply that it can communicate through several channels. The value is that those interactions can contribute to one organised owner record.

That is particularly relevant to hybrid landlords: people who self-manage some properties and use agents for others.

A good agent manages the property. The landlord still manages the investment, the decisions and their side of the relationship.

The Succession Question May Be Bigger Than the Efficiency Question

One audience member was approaching retirement and wanted to consolidate his information so he could eventually hand responsibility to his son.

That observation has stayed with me since I heard it.

We often sell technology through time savings. But for an established business or property portfolio, continuity may be more valuable than efficiency.

What happens when the founder steps back?

Where are the procedures? Who understands the history?

Which decisions were made and why?

What knowledge exists only in one person’s head?

A portfolio can be inherited in a legal sense while its operating knowledge disappears.

This is where knowledge systems, company “second brains” and structured AI support may create value far beyond drafting emails faster.

Our own agents COLLEEn and ALMa grew from that recognition. Some of the biggest business gaps are not tasks. They are knowledge that cannot be found, coaching that never happens and expectations that are not consistently explained.

The Collaboration Opportunity

The Guildford event reinforced my belief that practical AI adoption cannot be solved by technology businesses working in isolation.

Property operators understand the day-to-day reality.

Compliance and legal specialists understand the boundaries.

Technology teams understand what can be built.

Professional communities understand the questions their members are actually asking.

The best solutions will emerge where those perspectives meet.

We are interested in speaking with property-management companies, landlord organisations, software providers, compliance specialists, training businesses, SME networks and event organisers who want to explore practical applications responsibly.

That may involve a process review, a specialist agent or a route for testing Landlord PA with the right users.

The starting point is not, “How much AI can we add?”

It is:

“Where is the business currently losing time, control, knowledge or capacity?”

Bring us the part of your business that creates unnecessary chasing, delay or dependence on one person. We will help you decide whether the answer is a better procedure, a ready-made product, one focused AI agent or something more bespoke.

If you run a property business, technology platform, professional community or SME network and see an opportunity to collaborate on practical AI adoption, please get in touch [link to the contact page or Paul's LinkedIn profile]. We are particularly interested in speaking engagements, carefully defined pilots and partnerships that begin with a real operational problem.

My Question for Business Owners

If your business doubled over the next twelve months, which part would fail first, and how much of the answer currently lives in one person’s head?

Would it be communication? Compliance? Staff knowledge? Follow-up? Reporting? Approvals? Document control?

That answer is likely to be more useful than knowing which model currently leads a benchmark table.

Frequently Asked Questions

Do businesses need bespoke AI agents?

Not always. A better procedure or ready-made product may solve the problem. Bespoke development is most useful when a repeated, valuable business need cannot be addressed safely and effectively by existing tools.

Can a general AI subscription do some of this work?

Yes. General AI services are powerful. The distinction is that a dependable business system also requires procedures, permissions, reliable records, human decisions and monitoring.

Does AI replace staff?

Our approach is to give focused, repetitive or easily missed work to agents while people retain judgement, relationships, approvals and sensitive decisions. The objective is greater capacity and consistency, not indiscriminate replacement.

Can Essential AI speak to property organisations and business groups?

Yes. Sessions can cover our six-month deployment, practical starting points, human oversight, communications, knowledge systems or AI readiness for property and SME businesses.

Who is Landlord PA for?

Landlord PA is aimed at self-managing and hybrid landlords who want a more organised owner view across their properties, documents, actions and communications.

Paul Samuda is an entrepreneur and business and property mentor.