A bit of a different update today – as well as letting you know what’s new in Agilebase, we’ll look forward to how we see the year taking shape – we have an exciting roadmap for AI features and also events!

We think that 2025 will be the year of AI realism – certainly here at Agilebase, we’re working on separating the reality for the hype and making AI work for customers in practical, useful ways, whilst also taking into account valid concerns such as privacy and compliance.

We have a set of themes to bear in mind.

Themes

Context

What are the global trends? What developments are on the horizon?

There continues to be pushback against auto-enabling easy to implement but annoying functions such as summarisation and generation features. Even  large companies like Google have been doing this and it feels a bit like a modern day version of Clippy

Google has also recently reversed their original plan to start charging for AI features in their office apps. Instead they are raising prices across the board for everyone and making the AI features inclusive, removing the ability of people to opt out. Does that suggest confidence in their value?

That said, there also are plenty of really useful time-savers that people are paying for, e.g. https://fireflies.ai to do things like create tasks etc. from video calls

Technologically, experts believe that 2025 will be the year of model self improvement and agentic (i.e. using multiple different models interacting with each other) or chain of thought reasoning. What that means is cleverer AI models, so we should be able to find more genuinely useful purposes for them.

Privacy

We’ve already been working on this aspect as some of the earliest feedback from beta testers was their concern with where their data was going. Other customers have also mentioned that commercial contracts they have prohibit certain data sharing.

You now have the ability within your Agilebase organisation settings to specify a completely custom location for the AI your organisation calls, along with some other details such as which model to use, if multiple are available.

That means, while you can use a public cloud provider such as Microsoft OpenAI or Claude, for testing or non-sensitive/private data, for more privacy you can set up private hosted accounts, or ultimately host your own physical hardware. There are a number of GDPR compliant UK and EU companies who can help. Note the third party cost involved can be significant.

Those of you who came to our December Bristol & Bath AI meetup will have heard some interesting comments by Clive Kelman, former AI Principle Technologist at the Cabinet Office, about the UK’s approach to legislation, which aims to be more protective than the US, but also pragmatic.

Local capability

The privacy concerns above can often be mitigated by local hosting. Recent developments such as the situation at Tiktok  along with EU investigations into US companies have highlighted the potential fragility of relying on US infrastructure..

We have over the past couple of years been helping to develop local AI and tech capabilities and our ambition is to make Bristol and the West of England a world class cluster.

The UK has just been found the second most attractive country for investment behind the US so hopefully we have the wind behind us.

Some of the local companies we are  engaging with like Helix AI, who are coming up with really exciting innovations in the AI No Code space, including creating new open standards for how to integrate apps and platforms such as Agilebase with custom AIs.

The Bristol & Bath AI meetup also has a lot of great events coming up, which will help us all learn more.

https://www.eventbrite.com/cc/bristol-bath-ai-3454279

In February we are going to future gaze, we have Richard Godfrey and Rosie Bennett confirmed. March’ theme is governance, we have the wonderful Sue Turner OBE lined up.

Furthermore, throughout 2025 we’ll be involved with the Smart Cookies series of events at SetSquared in central Bristol. There are some really thought provoking talks coming up around themes of people and culture, innovation and money / finance.

https://www.meetup.com/smart-cookies-bristol/

Our Roadmap 

We aim to build a comprehensive roadmap of how we see AI adding value to the Agilbase platform. For now this is a summary of where we are and a look into the near future.

Current Situation

These are the main places AI is already used within Agilebase

• Creating new apps:  the AI is a way to really quickly build the basics – given any basic or detailed description, it will create a set of tables, add fields to them and connect them together
• Adding  simple or complex calculations to views: really useful for people unfamiliar with SQL. It can also troubleshoot and suggest fixes to calculations which have errors
• Running AI queries in an automated fashion: as part of the workflow engine, saving the results. So for example, if you have a list of products you sell, it can generate descriptions for all of them. In effect, running one chat query after another in a template based fashion, something like ‘Please write a description for product {X} given the details {Y} we have about it’.
• Generating documentation: In the background, the AI generates help for each part of your Agilebase app (in beta testing)

New out today

We’re really excited to be able to add to this list. You can now specify your own custom AIs to use, whether that be the company account with a public cloud provider, or an in-house private AI.

To do that, click your user icon at the top right, navigate to Profile then Organisation settings.

Enter the URL and authorisation key that your AI provider will make available and optionally, choose a model optimised for a particular purpose.

On the horizon

Improving AI workflows: 

Improving AI workflows so they are a lot more comprehensive and flexible.

Currently, you can generate a prompt to send to an AI and save the results to a field. However, to allow many more scenarios, we should allow the AI to create multiple records in Agilebase from one prompt.

For example, if you have your email system integrated with Agilebase, then you might have a comprehensive communication history for each e.g. sales enquiry. What can you do with this? Well an AI could generate a series of actions for internal staff, such as run a credit check, address a particular cost concern the prospect has and add a new feature request for something they’ve asked for.

To allow this we’ll be adding a new panel when editing a view, similar to the ‘workflow – edit records’ panel. It will be specifically for AI workflows and allow you to select

• the table to create new records in (e.g. a ‘tasks’ table in the example above)
• the fields to populate e.g. ‘task name’, ‘assigned to’, ‘priority’

As normal, you’ll be able to chain workflows together, e.g. one to assign tasks then another to create and send an email update to the prospect.

Helix integration

As mentioned above, Helix with their private AI stack are doing some innovative work on the practicalities of making complex AI applications accessible to all, which fits well with the Agilebase no-code ethos.

Two features of their offering are particularly relevant.

API integration: Agilebase can already create APIs for each view with the tick of an option. Using the https://aispec.org specification, Helix can automatically ingest these. What that means is that you’ll be able to ‘talk to’ the data in Agilebase. A simple example would be ‘tell me which purchase orders are outstanding’. That could be really useful for if you’re on a mobile device, or for users of large systems who may not want to search out exactly the view they need in the user interface.

Without using APIs, the amount of data you can feed into a prompt is limited to how large a prompt the AI will allow, which is typically relatively small.

You could also use APIs to create data, e.g. ‘please log a timesheet entry for me today for working on new AI features in Agilebase’.

Knowledge ingestion: we mentioned above that Agilebase can automatically create documentation for the entire system, containing details and the purpose of each view and table. Helix will allow us to easily interrogate that, so e.g. new users can ask ‘where can I find a report breaking down customers by country?’.

Other documentation like product info sheets or sales brochures could also be added for AI workflows to draw on.

There are many more possibilities. How and when we get to implement can’t be predicted 100%, but with additional help from people like Cardiff-based developer Ben Scattergood we are working to test and implement them as fast as possible. I expect new ideas will emerge as we go as well!

Pricing

Our pricing policies are taking shape – obviously the exact details will depend on which features we get working reliably and usefully. We only want to charge when we can offer great value, (e.g. time savings and improvements to customers’ own innovations and processes).

In the meantime, we’re looking for beta testers for early access to features, who will be able to try out new developments for free. Please contact oliver@agilebase.co.uk if interested.

Our aim is to be able to create pricing that will be a monthly addition to the Agilebase base price, plus a simple per-usage fee for higher usage levels, i.e. per number of workflows run. That way, costs will be straightforward, easy for business owners to understand and forecast. We can abstract away the more complex token and model based pricing of the underlying AI service.