AI General Thoughts on the Future, Anyone?

đź’¬ General Discussion
1

Salesforce is pushing their AI hard. Many of my SuiteCRM users are seriously considering spending tens of thousands of dollars to switch to Salesforce and the like for AI features.

Just looking for some general discussion around this. I can understand AI generated emails and responses could be good, but does anyone else know what kind of AI other competitive CRM’s are using?

I get the sense from talking with people that basically they think can eliminate their sales people! Use AI to create customized and personalized email campaigns and then respond to them and close the sale! I think this is a dream, but not a reality.

Are there any experts in the field in our community that could comment on what is real and what is pipe dream (at least for the foreseeable future)?

Another thing that I stress to clients is that AI is only as good as the data that’s in the CRM it’s learning from. If emails and call reports and opportuntiies, etc. aren’t been logged properly by the sales people, the AI won’t have the info it needs to base communications on!

2

This was a topic recently!

  1. Another fork of Sugar 6.5 Community Edition, called VTiger, from India, with a free open source version, has 4 AI features in its paid cloud hosted version, Calculus, Chatbot, Bot Manager, and Prompt Builder, which the latter one does the automatic AI follow up emails. Adding this AI feature to Suite would be a big time saver for users. Check it out:
  1. In the store there’s an AI Email Generator plugin which uses the OpenAI service to write sales and marketing emails, I’d give it a free 30 day try, to evaluate how well it could work for customers. After, it’s $19/month.
3

I agree with you Paul - rather than sitting down and defining specific requirements that might or might not be solved reliably by AI, customers seem to be driven by the general hype and marketing around this topic.

Can AI really help? In general, yes.
But where and how - and is your data ready / enough for AI?
I’ve trained a few ML models and even something simple like topic categorization / spam detection is completely wrong with 1000 records. From 10.000 records on, it starts to perform better.

And yes, AI is not going to automate sales any time soon, to fire the sales team. Rather, it can support the sales team.

Other than SuiteCRM, I’m working with Zoho in several projects and Zia is their AI solution.
There are features like:

  • When is the best time to call?
  • Churn prediction
  • Data enrichment
  • Anomaly detection
  • RAG (Retrieval-Augmented Generation) in analytics
    etc.
    which I see as valuable tools / features for sales and marketing teams.

One issue every SuiteCRM has for AI, is basically the amount of data.
If there are a few hundred leads inside your CRM, anomalies can’t be detected correctly, it’s just not enough data.
The SaaS apps have a huge advantage there, they could (not sure whether they do or allow them selves to do so in their toc) train models on all customer data.
It might be possible to train models on some specific jobs with general data - not sure about the performance though.

A good approach to implement AI in SuiteCRM might be to add it to the workflows.
There could be a logic like:
When a new lead is created via the webform and entered the website, open this website, write a 2 sentence summary and write it into a field Website summary.
(I’ve integrated SuiteCRM with Drupal as a customer portal - there is an interesting approach / module: https://www.drupal.org/project/ai_interpolator)

Automating email replies, well - depends on TOFU or BOFU.
I don’t think any general AI would be able to reply BOFU - that might be even difficult with a trained model on local data like chatd.ai and others.

Not speaking about GDPR in the first place - smaller companies in the EU might ignore this topic or don’t even know about details.
But in midsized and larger organizations, that’s a big no go.

Most of the business use at the moment seems to be questionable and being done without asking for consent when summarizing and transcribing phone calls and online meetings.
That would definitely be another great help for the sales team.

That’s an ease of use topic - it’s just so easy to hand all my (my customers) data over to Google, ChatGPT and others and they make my life so easy.
At the same time - at least in Europe - we’re at a huge disadvantage because of all the regulations.

So yes, the urge is there - clearly.
It would most likely help SuiteCRM in the marketing as well.
As for the CRM users, it would be interesting to have a list of day to day tasks and priorities to see where the AI potentials really are, in combination with an estimation of how low the fruit is hanging.
The VTiger prompt builder looks like a nice feature indeed. An implementation like this would probably require a bigger project.

4
5

Interesting discussion. I agree that AI in CRM should focus more on assisting users rather than replacing them.

We’ve have developed a solution SuiteAI Insight, with the goal of adding practical AI capabilities directly inside SuiteCRM.

Some of the features we’ve found useful in day-to-day CRM work:

AI Email Assistance
Inside the SuiteCRM email composer, AI can help:

  • generate replies

  • improve or rewrite emails

  • adjust tone (professional / friendly)

  • translate messages

Prompt Builder / Custom Agents
Admins or users can create prompts such as:

  • meeting briefings

  • opportunity summaries

  • follow-up email generation

These prompts use context from the CRM record and related activities.

Multiple AI Models
The system can work with different providers like:

  • OpenAI

  • Gemini

  • Grok

  • Local LLMs

So organizations can choose between cloud AI or self-hosted models.

7

Am I the only one who thinks the replies here are landing in the same place? I read through them, and yeah, I’m with the “AI as a sales sidekick, not a salesperson” take. The overlooked bit is all the CRM busywork - digging through notes, emails and meetings before a call.

The data-quality point still matters, though. If the CRM’s full of stale contacts and half-baked notes, AI can make the mess look polished, but it’s still a mess.

8

About the busy work:
Most of the AI ideas that I’ve seen so far, are quite developer centric. Mostly a proof of concept.
Most likely, some might be further developed, there might be more feedback from real world users in production day to day tasks and the AI topics will gradually change to real world solutions.

You’re perfectly right about the data quality.
If it is low, AI will just produce more low quality data.
That is not an AI problem as such and exists for as long as software with human input existed.
The topic is just not as fancy and as hype as AI, so the focus shifted.
As soon as the AI flattens out, people will see other topics again and hopefully data quality, processes and CRM maturity are among the new old topics that are visible again :wink:

For my projects, I have so far developed 0 product like solutions to advertise here and mostly spoken to the users, discussed and reflected on their needs, their ideas and pain points and have quite some specific AI processes in place.
So far, I don’t see a one size fits all AI product in or on top of SuiteCRM. But plenty of room for specific edge cases and parts of day to day business processes.

9

I think in general people a are “de-Saas-ing” (I just made that up). Lots fo companies are putting considerable effort into building their own in house apps by vibe coding. I think for the foreseeable future Saas is not a good business model. Clients will just code their own.

Eventually, I think the pendulum will swing back the other way. When clients realize the scope and feature creep of thier internal software, security issues and maintenance costs, Saas will look more attractive again.

Right now the math is” You’re paying $x a year for Saas when you could just vibe code it yourself, thats a huge savings”

Tomorrow the math will be” You’re spending all kinds o money on developers, tokens and hosting environments as well as significant maintenance costs to run your own in house software… Saas can do it cheaper and remove the the headaches”

10

Yes, I completely agree with that.

Most companies which move now to vibe code are fairly unhappy with their current SaaS or custom developed software and hope that AI code will solve all their issues.
Once they realize that producing more code isn’t the issue, they’ll have to re-think (or go for ignorance is bliss).
The big question is then: When will we see budgets being freed from going back from vibe coded software? This could be a massive one way street.
Who wants to do the data migration from a vibe coded something into a stable and standardized system?
Wondering already, how to explain this to the non technical vibe coder decision maker. :wink:

I’m in multiple projects where some parts of the infrastructure is vibe coded. I’m very happy to be able to observe this first hand.
So far, two customer portals have been way more expensive than building them on Drupal and they’re not integrated and feature creep is growing fast.

11

I don’t think vibe coding itself is the problem. The real issue is when someone without enough experience or understanding of SuiteCRM tries to build or modify things using AI. If someone already have solid SuiteCRM experience, AI can be a powerful tool to speed up development and solve complex problems. The combination of experience and AI can be a real game changer.

For example, a customer might ask AI to implement a feature in a module, and the code may work perfectly today. But if the changes are not upgrade-safe, a future SuiteCRM upgrade could break the functionality or even cause bigger issues.

12

We had an example of this recently to highlight this.

We had a call from a client (we basically do their website), they have built an “Engineer Workspace” can we host it for them. Asking what it was built in the response was “chatGPT” - I obviously asked from more info.

Turns out someone (not sure if it was an internal person or had been contracted out) had vibe coded this app - details were provided, basically it was tied to Cloudflare services.

The key point was that they wanted us to “host” it and provide administrative support, we had to “decline” this opportunity, their manager got in touch with a ream of an email basically saying that we had mis-understood and maybe a meeting would resolve this.

I had to explain that we couldn’t host it (Cloudflare would do all of that and they can obviously deploy it to CF as they already had it working there), we couldn’t admin it nor support it (it used language I hadn’t even heard of). Not too mention that have had an AI generate thousands of lines of code to produce this with no warranty (provided to me at least) nor documentation on deployment, code structure etc.

I am quite ready and willing to help our current (and future) clients with hosting packages that have existing support, forums that can be used to search for help, by vibe coding you are effectively building Custom Apps every time, you are really tied to your original developer for support and can only hope they have inspected the code base and understand it.

AI (well what’s currently called AI anyway) should be used as a helper tool, trawl through vast data pools and find opportunities, inspiration, ensure you don’t miss things and generally act as a gopher :wink: certainly not replacing staff (its definitely not there yet).

Mark

13

Good point I hadn’t considered what would happen if one agreed to host and support a vibe coded app on their sever.

14

Quick add-on: I found Sugarbug recently, and it’s made pulling context from different tools much easier. The data-quality point still stands, though - AI can’t fix a messy CRM.