Weāve seen AI on the rise now for the broad society for over two years now.
Itās been a huge hype topic, a lot of pros and cons. Plenty of great use cases in so many different areas.
Iāve spoken to many clients, friends and other users and learned so many interesting cases, theories, expectations etc. - a great time, with plenty of conceptional shifts.
Iām curious about the relationship to business software / CRM and SuiteCRM especially:
Weāve seen the recent drop in stock prices for Salesforce, SAP and others.
āAI is eating softwareā ā¦
Whatās your take on it?
Where are we heading?
Do we need CRM in the near / mid future? Or just agent teams with access to our DB?
What is your company doing? Did you leave / decide against (Suite)CRM with AI in mind?
I think we are heading inevitably towards world where AI is used everywhere.
This changes things, but it also opens a whole bunch of opportunities for projects like SuiteCRM.
We have open platform, which has a long history and a lot old code. The structure is mature, but changing things require a tedious task of research in code, a comprehensive testing and then you can provide the changes. All things that AI can help a lot.
We need CRM in future, in form or another. But we should really start thinking how to evolve faster. The view for business data should be as complete as possible. The default SuiteCRM deployment should be faster, easier and more complete.
If all SuiteCRM users would combine their effort somehow, what could we achieve? How could we have a platform inside the production CRM for sharing ideas and fixing problems? What if we had a support channel inside SuiteCRM, maybe even AI driven?
The futurists are saying we no longer need a CRM, just a database with an AI front end. I can see how thatās possible, but not functional or plausible. Fundamentally CRM is about ārelationshipsā between people and companies. AI can definately help with data mining but fundamentally, you need a person that deals with another person to get most business transactions moving (Iām talking B2B here, I get there peopleless transactons like financial transactions and B2C ecommerce that doesnāt āneedā people").
I want to āseeā the data, the last call report, what else is going on in account easily and in a dashboard that tells the story. I can get a āfeelā (that AI canāt get from pure data) on what the next best approach is to the person or persons making the decsion. I get AI can summarize this and that is sometimes useful, but I still see the need for a dataāhuman interface (a CRM!).
What would pure AI CRM look like? I guess I can envision alot of the things it already does. Create call and meeting summaries, ID contacts and companies based on data/trends and targetting that would be high priority leads, maybe even start the outreach process with cold emails. These are all great!
At some point (for most B2B anyway) a human being is going to have to connect with another human being and one human being is going to have to trust the other human being enough to sign a contract.
AI can definately help with that, can even summarize the meeting, prep the demo, write the propsal (all great uses of AI) but unless the future holds that AI is making the buying decision and I guess my AI could just talk to your AI. If that doesnāt happen you still need a human or two in the loop. That human is going to need a CRM to organize and display data in a human way.
A future proof CRM must be that graphical and usable interface between the data and the person.
The future seems to be based around AI as a Core and that handles all the requests, being the Front End for all the data/records which are being used. Just like the Comet Browser, there would be Visual Interface to show records, draw reports or notify the users, but the core backbone is apparent to be Based on LLMs. With Manual Call Handling, Email Messaging or Reminders setting going away, AI would take over most of the āAdminā stuff.
With Customer Services Domain disrupted, health care and finance are what the AI Companies are targeting now. The Trend to develop custom softwares has greatly reduced ecp for New Custom Softwares. Those who are still running the show have a moment of their past connections or a product they have been working on for so long.
On the contrary if AI Subsidies are reduced, the actual cost of Using AI would skyrocket, thus demanding companies to utilise a human resource instead of the Costly Agent for solving an problem.
For now, AI is a companion for boosting Performance 10x. Adapting it has become necessary and we follow wherever the wind blows.
Great thread . We all love technology, and we all love humans in this group, I would say. Thus, a perfect combo to āget AI rightā.
I think everyoneās points above are fair, but Iām very hesitant to trust a bot with sensitive customer data, for one, but if that can be dealt with, and everyone be happy with that part, then Iām open to consider things. For example, Iāve already started dabbling with letting N8N do something with an API so really, probably thatās not any saferā¦but at least you can probably control the level of permission on a granular level with APIs.. but I digressā¦
What I would envision, if AI were used, would be to āenhance the human experienceā, not replace it. So, what I would see is something like this:
Normal, full visibility dashboard, GUI - for all the reasons mentioned above.
* Prompts to follow certain actions, that subsequently contain useful automations, with the following example:
start typing āJohn Doeā
āDo you want John Doe from ABC corporation, or from DEF Ltd?ā|
āDEFā
āGot it . You are logged in as a sales rep so would you like to look up your last task and notes you have on John Doe from DEF Corp?ā
āYā
āSure, hereās that task. Would you like to communicate with him?ā
āY. Emailā
āSure, hereās a blank email. Do you want to build the email from a template? You have a few templates with customizable blocks in them.ā
āy. long time no talk emailā
āSure, hereās the customizable āLong time no talk email set up and ready to customize for Johnā
Click send
āOk thatās sent but do you want to set up a follow up on this?ā
āY, 90 daysā
āDone. You will get an email from SuiteCRM in 90 days when itās time to check in on John Doeā
Something like this would be awesome in my opinion and keep us in more control of our data (I think?) while taking advantage of āenoughā AI?
Definitely I would not hand full control of our customer data over to a bot without having crazy strict guard-rails.
I donāt see CRM disappearing anytime soon. AI is useful for summarizing calls, finding stale leads, suggesting follow-ups, and preparing reps before a conversation, but someone still needs to verify the context and own the relationship. Iāve found dashboards and activity history much more reliable than asking a chatbot to interpret everything from raw records.
CRM will always provide the core data, data relationships and notes until AI memory becomes all encompassing. While probable at some point, at some point there will be push back on privacy.
For us suite has become a centralized, extendable managed database in addition to ācoreā crm functionality. 70% of our UI interactions, especially on new record set up and enrichment are done via Claude chat and make.com mcp, wired scenerios, while the UI is 15% traditional and 15% in studio extending to make the structures better support the needs of automated workflows, ie record states, new modules like capturing all our social content for review in the suite UI, then republish to web. Suites extendablity is an easy way versus building new data structures elsewhere and the UI comes for āfreeā. My suite dev has also launched an MCP for suite that seems pretty exciting as wellā¦looking forward to the demo.
Iāve been doing some cool stuff with AI personas lately. For example, I built a virtual customer.
Iām re-designing a website for a client, and I took everything I knew about her and her likes and dislikes and created an AI Wendy. I then gave AI Wendy website design and taste skills and had AI Wendy build her own website. Then I built a learning loop where Real Wendy was giving feedback to AI Wendy. AI wendy updates here knowlege about what the Real Wendy likes.
So far this is working out spectacular. The Real Wendy couldnāt be happier.
You can read more on my blog if you are interested.
But lets bring this concept back to CRM. CRM is a data warehouse and the interface is essentially trying to allow the user to gain knowledge from the data. All the AI implementations so far are quite superficial.
Summarize a meeting, lead research, write an email, do some fancy data analysis.
Seem cool but not really helping increase sales.
What if we extended my Wendy concept to CRM where every contact had a Knowlege layer. The knowledge layer would be things like, ānever call on fridays too busyā, āAlways asks for a discount, quote highā . The cool thing is that a person doesnāt need to populate this. The knowledge can be distilled by AI. Every email goes through an agent to comb it for any strategic account information or contacts perferences. Same with meeting summaries to update the knowledge layer. These could be simple .md fields in a āKnowlegeā module.
Then the CRM is no longer a source of data. Its a source of knowlege. If a rep leaves no problem. All the account and contact knowledge is stored in the CRM and a new rep doesnāt have to review the history to figure it out.
You could also us it pre-emtively like āwould Wendy accept this quoteā. How would Wendy feel about how we resovled this case.
You can also ask questions of AI that has access to the knowledge module. It has the summary so it doesnāt need to read ever email and review every quote when you ask it a question.
You can make the CRM 1000% more valuable this way.
Iām working on buidling this, itās not really that hard. It just an extension of what I learned building AI Wendy.
check out the ideas behind Hermes Agent USER.md and MEMORY.md - thatās exactly it (just for users).
Essentially youād be building this, based on leads / contacts interactions.
To optimize the idea for a CRM, Iād prefer to have the result in structured fields:
Best days to call (multi select automatically filled)
Other CRMs pre-calculate this and show the next best time to call directly in the contact.
Price sensitivity (dropdown field).
The challenge here is, compared to a mark down structure inside a text area (actually, this would be a cool feature - mark down support, beside the WYSIWYG/HTML field), itās opinionated and usually defined by the company / business processes which fields are required for segments etc.
Your approach with the MEMORY.md or PROFILE.md or similar works well for AI and humans who want to read through this.
Iād somehow try to add options to use the resulting information in reports, segments, workflows etc.
Currently Iām working on quite a nice project - the customer has over 18 years of data generated by over 150 sales reps and most of it is unstructured in notes / description fields.
With the migration to SuiteCRM, weāre moving all this information over, but weāre trying to accomplish the opposite and generate structure out of this mess.
As for the AI benefits: If there are structured fields, your AI enabled processes get more deterministic and much faster & cheaper as well. In most projects which reach a certain level of maturity, there are only a few dozen fields which help sales & marketing with structured data. A mix might be great: Add the MEMORY.md into a module with your approach, additionally, try to extract data from there for some predefined, more structured fields to make even more use of this feature?
I get structured fields are easier to sort and filter on, but this will be dramatically different for every client/contact and they may have to really create a bunch of fields and maybe never use them.
My concept is maybe 3-4 large .md fields that can be structured like an .md file. The Agent only needs to read the .md file fields for context.
Another agent reads emails/meetings/call reports and updates the .md fields with knowledge.
Maybe an agent that runs at night that takes the days data and distills it into longterm memory and also consolodates the .md files so they dontā get out of hand.
I donāt anticipate these being thousands of lines long per account/contact. They are not meant to give wide context like ātell me everyone in the manufacturing industryā. That a field definately works better as a field.
For questions like, what is the best approach for Dave. Is definately an .md file and not a field.