AI chat is the new admin console for business systems
The short answer
The point-and-click admin screen is no longer the best way to configure a business system. Asking an AI to make the change is faster and more accurate, whether the system is Salesforce or custom software. The same is true for reports: instead of waiting for an admin to build a dashboard, people can ask for the numbers or the chart they need. Chat is becoming the interface for getting information, changing the system, and getting work done.
How configuration used to work
I implemented Salesforce for companies for more than fifteen years. Salesforce is designed so that a lot can be done with point-and-click setup instead of code. That is true, but it is not easy, which is why there is a whole industry of consultants and admins. Anything that could not be clicked was written in code.
The setup screens are slow. Renaming or deleting a field can take half an hour once you trace where it is used. Reports and dashboards take longer still, and tools like CRM Analytics and Tableau are powerful but complicated to configure.
How I configure systems now
Today I do the point-and-click work with Claude Code, even on Salesforce:
- Objects and fields
- Permission sets and profiles
- Page layouts and list views
- Validation rules and automation
- Reports, dashboards, and custom components
I describe what I want in plain language. The AI makes the change in the code repository, and it is deployed like any other change. It is faster than the setup screens, and more accurate, because the AI does not forget the fourth place a field needs to be added.
For dashboards, I no longer use the built-in dashboard builder or a separate BI tool on Salesforce. I have the AI build a custom component that displays exactly the numbers the team needs, and it looks better than the standard options.
What this means for custom systems
This changes what a custom system has to include. Platforms like Salesforce carry a huge configuration layer: screens for defining objects, fields, layouts, permissions, and workflows, because they are built for every company and every company configures them differently.
A system built for one business does not need to rebuild that layer. New fields, rules, workflows, layouts, and permissions are changes to the code, made by an AI coding tool and deployed through testing. The admin console is a conversation. That makes a custom system simpler to build and easier to change.
What this means for the people using the system
The same shift applies to everyone who works in the system, not just admins.
The old way to get information: request a report, wait for an admin or analyst to build it, then find that it answers almost the right question. The new way: ask. "Which properties had the most support cases last month?" "Show me renewals due next quarter by account manager." "Chart signups by week this year." The answer comes back as a table or a chart, from live data, in seconds.
People can do more than ask questions. They can update records, draft messages, and complete routine tasks from the same chat.
The guardrails that make it work
Letting AI change a system or read its data only works with the right controls:
- Changes go through version control. Every configuration change is in the code repository, with history and the ability to roll back.
- Testing before production. Changes are tested in a separate environment and approved before they go live.
- Permissions apply to AI too. When someone asks a question, the AI sees only what that person is allowed to see.
- Actions are logged. What was asked, changed, or done is recorded.
These are the same practices professional software teams already use. They just now apply to the AI doing the work.
Where this is going
Setup screens, report builders, and BI configuration were the interface because there was no better one. For a growing share of the work, there is now. Systems designed with that in mind are simpler, faster to change, and easier for everyone to use.
Where Opero fits
Opero designs, builds, and operates custom business systems, with AI built in for the admins who change the system and the people who use it. See custom CRM development, or start with a free review.
Common questions
Can AI configure Salesforce?
Yes. AI coding tools like Claude Code can create objects, fields, permission sets, page layouts, flows, reports, and custom components through Salesforce's metadata and a code repository. It is usually faster and more accurate than the setup screens.
Does a custom system need an admin interface like Salesforce Setup?
Not a large one. Configuration changes such as new fields, validation rules, workflows, and permissions are made through an AI coding tool working in the code repository, then tested and deployed like any other change.
Is it safe to let AI change a business system?
It is when the guardrails are right: changes go through version control, are tested in a separate environment, and are approved before production. For questions about data, the AI works within the same permissions as the person asking.
Do we still need dashboards and BI tools?
For the handful of numbers the business watches every day, a built-in dashboard still makes sense. For everything else, asking for the report or chart you need, when you need it, is faster than building one in a BI tool.
Founder of Opero. Building software since 1998 and consulting since 2008, designing, building, and operating business systems.