Subscribe to our newsletter to receive business tips, learn about new funding programmes, join upcoming events, take e-learning courses, and more.
Take the first step to successfully starting and growing your business.
Join for freePosted: Wed 8th Jul 2026
Most small businesses already sit on useful data – in spreadsheets, accounting tools, CRMs and website analytics – but struggle to turn it into clear answers.
In this practical Lunch and Learn, Harinath Selvaraj shows how AI-assisted reporting can help you ask plain-English questions and get charts, KPIs and next steps – without hiring a data team.
Learn why many SMEs are "data-rich but insight-poor," how to use AI assistants safely with your own business data, and what to look for when choosing tools.
The session includes a live walkthrough of asking business questions in everyday language and getting visual answers, plus practical steps to get started (including free options).
Why most small businesses struggle with reporting: Data stuck in spreadsheets, no single view of performance and what's actually driving sales or costs is left unanswered until month-end (or not at all)
How AI changes the game for everyday reporting: How to build charts and dashboards quickly, track KPIs and ask analytical questions in plain English (e.g. Which products sold best last quarter?) without SQL or a dedicated analyst
A practical path to get started: A simple workflow to upload business data, create a live dashboard and use AI-assisted analytics safely – including what's realistic on a free plan vs when a paid plan makes sense as the business grows
Harinath is the founder of Granola Consulting, where he helps businesses adopt practical AI through product modules and bespoke delivery.
After more than 15 years in data science and analytics – including six years as director of data science at Hertz, where he built and led a global team and delivered large-scale ML and Gen-AI programmes – he now focuses on making enterprise-grade AI accessible to SMEs.
Watch more expert webinars
Access a growing collection of expert-led webinars covering marketing, sales, finance, growth and more – ready whenever you are.
Lightly edited for clarity.
Ryan: Good afternoon, everyone, and welcome to today's Lunch and Learn. My name is Ryan, and I'll be your host today. For those of you attending Lunch and Learn for the first time, Enterprise Nation is a vibrant community platform for start-ups and small businesses.
Today, I'm really pleased to introduce Harinath Selvaraj, founder of Granola Consulting. In this session, Harinath shows how AI-assisted reporting can help you ask plain English questions and get charts, KPIs, and next steps – all without hiring a data team. As always, if you've got questions, post them in the chat, and we'll do our best to answer them at the end.
The webinar is recorded, and as always, a follow-up email with the recording will go out later today. So on that note, I will hand over to you, Harinath.
Harinath: Thanks, Ryan. Good afternoon, everyone. My name is Harinath Selvaraj. I am the founder of Granola Consulting, an AI consulting firm based in Dublin, Ireland.
Prior to that, I was a senior director of data science for Hertz Car Rental. I have around 15 years of experience working with products and productionising them in bigger markets like the US, where this resulted in around 25 million in revenue uplift. There are a lot of things you can do with AI nowadays, and I'm going to talk a bit about the revolution at the beginning, then go over some of the key challenges that SMEs face with reporting. This includes figuring out where they are as a business, what areas they could improve on with the help of AI, checking KPIs, and making sure business performance is well understood and in a shape that helps business owners take strategic action.
So, about me – I've already mentioned 15-plus years in data science. Currently, I help SMEs develop AI products that have an impact on their business performance. One of our flagship products is HoneyGold, where we provide analytics for SMEs, using tools like Claude to help answer business questions.
As many of you attending will know, businesses have changed drastically over the past several years with AI. It's estimated that around 70 per cent of Irish workers are currently using AI, and this number was just eight per cent two years ago and is rapidly growing. In the UK, it's around 73 per cent of workers who use AI.
So the point here is that you're not really competing with AI, but with people who know how to use AI in their day-to-day jobs. As a business, it's now really easy for anyone to copy what you do and try to provide better services with the help of AI solutions. So it's on us to take advantage of the latest advances in AI and put them to work for our business use cases.
The key problem I wanted to highlight is that small businesses are often data-rich but insight-poor. There's lots of data that small businesses collect – one source is sales, another is marketing, a third is customer transactions, and a fourth is company performance, such as employee pay. The main problem is that since this data lives in multiple places, it's really in silos. Getting an answer to a specific analytical question takes a lot of time because you need to combine sources and slice and dice the data to get where you want to.
So what businesses typically do, rather than hire a data team, is keep their data in Excel spreadsheets and build some kind of accounting dashboards and charts. If the data grows at a rapid pace, they might have systems to pull CSV data out of individual systems and merge it into a single Excel sheet for reporting. This is only possible if you have, say, 100,000 rows or so – beyond that, it becomes extremely difficult, because tools like Excel are only suited to handling a certain amount of data. After that, the computer struggles as the data volume grows.
Luckily, what we have now is AI for everything. If people have any doubts about whether something is possible, they can upload a document to AI and ask questions directly. There are lots of tools, and I hope the audience here knows about Claude, which is quite popular, along with Gemini, ChatGPT, and Perplexity. These are probably the famous ones, and at least one of them is likely your go-to tool for everyday use.
I'm going to show you, in the next few slides, how to integrate analytical tools with Claude. The major problem with these AI tools is that you can only upload a handful of data – each tool has a limit, say 50 MB, or 100 MB in the case of Claude. So you have to be really creative, uploading small batches of data and combining them in Claude or Gemini to help answer questions.
Even then, these datasets aren't persisted to the servers, so every time you have to update the data locally and re-upload it to get an answer. It's a tedious process once the data volume grows, and that's where scalable solutions like BI tools really favour big data analysis. That's exactly what I'm going to cover in this slide: comparing different BI tools.
If you already have some knowledge of business intelligence tools, you'll know about Power BI and Tableau, which are quite famous. But these are enterprise tools, only really suited to bigger customers like corporations and multinationals who want to process billions of rows of data in real time and push it into an analytics layer for reporting. What we developed instead is HoneyGold Analytics, which we provide for 25 euros per month, with a 14-day, no-credit-card-required free trial, and you can cancel at any time.
It's similar to these other tools, but what really differentiates us from enterprise-grade tools is ease of use for SMEs. SMEs typically have little to no understanding of how business intelligence tools work, and they need a lot of hand-holding at the initial stages, which big companies don't offer smaller businesses for budget reasons. That's where we come in – our analytical tools can plug directly into tools like Claude to answer business questions, which I'll demo in a bit.
Okay, let's go to the demo. For this demo, I'm using UK house prices because I thought it would be more relevant for this session, given that a lot of the audience is coming from the UK. This dataset is obtained from a UK government website, and there's a link to it. It has around 100,000 records, totalling around 125 GB.
That's a decent-sized file for testing purposes, and what I'm going to do is show you how you can use Claude to answer questions about this data. So, the link I mentioned earlier is from gov.uk. You can get price-paid data, which is the exact amount paid by customers buying a house in the UK housing market.
From there, you can access lots of data – this is one big file containing all housing transactions from 1995 onwards. For this demo, I'm just looking at early sales data from last year, which comes to around 130 megabytes. I'm going to show you how I've uploaded it and integrated Claude to answer questions.
Before that, if the same data had to be imported into Excel, depending on the client machine's size, it would take a long time just to load the values. Even creating a single chart would take ages, because of the computing restrictions on individual machines. So what I'm going to show you is how I've uploaded that data to HoneyGold Analytics.
If someone has purchased a HoneyGold subscription on the Granola Consulting website, they can sign in here. This is the signed-in web page, where you can create dashboards and charts. This is one of the sample dashboards, as you can see, with sales data.
I've also created charts for the UK house prices. This particular chart shows the number of houses sold for each county, and I think Birmingham stands at the top, with around 12,500 properties sold last year. For people who have little to no idea how charts and dashboards work, you can use Claude for creating all of this instead of relying on this interface.
It requires a bit of learning to understand how the charts and columns work – there are different fields that you can drag and drop. But instead of doing that, what I've done here is use Claude to help answer questions about this data. So this is Claude, and I've got a paid subscription – around £20 per month, though you don't have to get it if you're not going to use it much.
What I've done here is add HoneyGold Analytics as a connector, and now I can ask any question I want about this data. So the first question I'm going to ask, since I don't know how many datasets there are, is: "What are the different datasets available?" Now, since it's already connected to the HoneyGold Analytics plug-in, it's going to directly call the integration, and now it's showing me all the different datasets.
It also provides other information – for example, this dataset was uploaded 10 hours ago, and it looks like it's coming from the Land Registry property dataset. If I want to see the rows and columns in this dataset, my next question would be: "Can you provide sample rows from UK property sales data?" The beauty of AI is that it's so good at understanding the questions you're asking.
Even if you miss certain key information, it automatically tries to match what you've said with what it already knows. In this case, I used the term "UK property sales" – not the exact dataset name – but since there are only three datasets in total, the name I provided closely matched the one I meant.
So this is the sample data you can see: different properties that were sold, the postal code, and the county and district where each was sold. A "new" field denotes whether it's a new-build or an established property, and the property type could be terraced, semi-detached, detached, bungalow, or apartment, and so on. Now that we've got sample data, every business will know, to an extent, what data they have.
I'm going to add a glossary for the sake of time. This is the glossary for this data – I've added more detail on each field. For example, "date of transfer" is the date the sale was completed, as stated on the transfer date, and "property type" should be detached, semi-detached, terraced, flats, and others.
Now I'm going to send this to Claude, and it says that's a helpful reference. So now I'm going to ask five sample questions with Claude, and the point here is how easy it is for SMEs to ask questions directly to Claude. It also shows how quickly Claude does the calculations behind the scenes, fetching the data and providing accurate results.
The first question I'm going to ask is: "What is the average sale price by property type?" I think it shows the average price for detached houses is around 500,000, and there's a handful of other properties I'm not sure about – probably big mansions or bungalows that come with a huge piece of land – sold for around a million. Semi-detached, as you can see, sold for slightly less than detached, and flats are around the same price range as semi-detached.
I think the reason is that apartments in London and other metropolitan areas contribute to the higher average price for flats. The next question I'm going to ask is: "How many properties were new-build versus established properties?" It says new-builds were only five per cent of total sales made last year.
I believe the government has been talking about increasing the number of new-builds, which should give a better figure this year. If you had even a few more months' worth of data, this number could be a bit higher than last year's. So, considering these general numbers for England and Wales, my next question could be: "How does the distribution look for London?"
I'll ask for a pie chart, because it's easier to visualise the numbers than to read them from a table. Let's see how Claude responds. So it's provided a table showing 4.4 per cent were new-builds, and a pie chart showing new-build versus established properties in the London housing market.
Now, if I wanted to see how many new-builds versus established properties were sold last year for each city in the UK, I could ask a follow-up question. This time, I'll ask Claude to provide a bar chart with different cities, showing the proportion of new-builds versus established properties, with the highest new-build proportion listed first, followed by others with the lowest proportion. As you can see, I think Salford has the highest number of new-builds.
Since I don't follow the UK housing market closely – and I know Kumar is watching this – please let me know in the comments if these numbers make sense. So, the point here is you could ask any number of questions you want with Claude, and our HoneyGold starter subscription provides 2,000 credits for asking these questions. That works out to around 50 questions per day, at a fraction of the cost that big enterprises pay for native AI integration.
I'm sure there'll be many questions about how this all works and how to perform the integration, but everything is on our website. If you'd like to sign up, you can do so now. It's a monthly subscription of 25 euros, and it also offers a 14-day trial for new users.
I'd like to finish by saying that a lot of businesses have already started using AI in their day-to-day work, and if we don't use it, we'll probably lose market share. It's not too late for businesses to adopt AI, and the faster you do it, the more secure your business and market position become. That's pretty much it, and I'm happy to answer any questions you have.
Ryan: Thanks, Harinath, that's really interesting. It's amazing what you can do, isn't it – all that data, and how quickly you can do it. Thank you for the demo. I've got a couple of questions.
How clean does the data need to be before you upload it back into Claude?
Harinath: It depends on the business. Nowadays, the mantra is to use AI for everything, so even for cleaning data, you could use AI. But you have to put certain practices in place to make sure you provide clean data to AI, because it's always said that AI isn't a magic tool that can fix everything, though it can try its best.
And it comes with a cost too – as you know, Claude Fable has been quite expensive to operate. But you can pretty much do things on your own, or ask a consultant to help clean the data. The core point is: garbage in, garbage out. If you cannot provide clean data, you will end up getting inconsistent results, which is bad for your business.
Ryan: Yeah, very good point. I think that's it – like you said, the quality of what you put in is going to affect what comes out, isn't it?
Harinath: Yeah, good point. Brilliant.
Ryan: And how often should businesses be reviewing these AI-generated reports? Should it be daily, weekly, monthly, or does it really depend?
Harinath: I think it depends on how different businesses operate. For example, smaller businesses concentrated on marketing or sales might need to look at daily numbers to see how they're performing, how customers have come in through the funnel, and make any modifications before they start losing customers. So it depends on the business's urgency.
In bigger companies, they have something called KPIs – key performance indicators – for every smaller team, to measure how well they performed. For example, a customer care team might build KPIs around how many customers called customer support, how many got a resolution, and how many didn't. Other KPIs might include what rating customers gave, how many pieces of feedback were received, and how many customer escalations happened in a day. It definitely helps to frame all these KPIs for the business, but it's also on us to really understand, as business owners, what we're trying to solve and how to take advantage of AI to help along the way.
Ryan: Fantastic, thanks, Harinath. We'll just do one more question, since we're on time. Leo was just asking: how does the system help boost business revenue? I guess that's the main thing everyone's thinking about, isn't it?
Harinath: Yeah, right. So, again, tying back to the KPIs – if you have a certain business, let's say even a flower shop, and you wanted to increase sales, you'd end up doing marketing, making sure your digital presence is well established. You'd have to look at improving your website content and how to use SEO effectively.
AI plays a big part in this – nowadays, people don't do Google searches like they used to a few years back; now they directly ask tools like ChatGPT for everything. So making sure your business is listed on these tools, whenever customers search, is also important. Making sure you produce original content is also very important.
What's your moat, what's your unique selling point, and how is your customer experience far better than your competitors – and how are you able to show that to your audience? If 10 customers go to your shop and have a wonderful experience, think about how to make your customers become your brand ambassadors. You could do this by asking them to provide a review on your Google Business page, or even having a session with them to talk through their experience.
This applies to marketing agencies or bigger companies wanting to establish trust. There's a whole lot involved, and I believe a lot of you have already been using Canva for posters and so on. There's AI built into all these tools, so you can ask, with a prompt, for the kind of design you want.
Try to make use of AI as much as you can, and learn every day, so you stay on top of your business, because competitors are fiercely working towards beating the champions. So it's on us to really allocate budget and upskill ourselves in using AI, and make sure it delivers what it claims to. You're not learning AI just to get awareness, but to put it into use and see how it generates value for your business.
Ryan: Amazing, really helpful, Harinath, thank you. We're just a bit over time, so Leo just said, "Great session, thank you for sharing," and I think it's been really interesting. As I said earlier, the recording will go out later today, so if anyone wants to watch back or connect with Harinath, I've popped your Enterprise Nation link in the chat, so please do reach out.
I know we never have enough time on these sessions, but this was really interesting. Thanks to everyone for joining, and thank you, Harinath – it was a really, really interesting session.
Harinath: Thanks, everyone. Thanks, Ryan, for the opportunity. I really appreciate it. Have a wonderful rest of the day, everyone – enjoy, and stay cool as well.
Ryan: Bye-bye, Harinath.
Unlock more on-demand sessions designed to help you sharpen your skills, grow your business and stay one step ahead. Find more Lunch and Learn webinars
