April 9, 2025
Is It Time To Replace Your System Admin With AI?
Let’s start with a confession. At Brimma, we recently caught ourselves doing something hilariously old-school: setting up a knowledge base for our support team. Like it was 2013 and Clippy was about to chime in.
But then came the epiphany: Why build a static, aging help archive when we literally develop AI tools that can read, learn, and advise in real time?
So we pivoted. We plugged our product release notes into our own AI, trained it to prioritize the latest updates, and created a digital support rep that’s always on and always up to date. It saved us time, streamlined onboarding, and created a smarter, more agile support desk—before we even hired our first agent.
A Lender’s AI Revelation
Around the same time, we spoke with a client who managed a team of LOS system admins whose jobs were to maintain user access, troubleshoot issues, roll out new configurations, and keep the entire mortgage tech stack humming.
The twist? This leader had never been a system admin themselves, and as a result, had no real way to vet whether the solutions their team proposed were optimal. That’s when they got creative.
They built a personal GPT. They fed it everything—internal documentation, release notes, admin guides. What they ended up with was more than just a searchable knowledge base. It was an expert assistant capable of proposing solutions, explaining implications, and helping them collaborate more effectively with their team.
The result?
- More work completed per admin
- Better design decisions
- Faster adoption of new LOS features
- And a leader who could finally lead confidently
So Why Isn’t Everyone Doing This?
Simple: most lenders are still trying to operationalize survival, not innovation. They’re so caught up in reactive firefighting—user issues, release surprises, compliance updates—that they don’t have the headspace (or leadership mandate) to rethink the admin role altogether.
Add in the myth that “AI is only for IT” or “it’s not ready for real work,” and suddenly the status quo feels oddly comfortable.
And let’s be honest: many LOS vendors aren’t rushing to simplify things either. Complexity keeps their platforms sticky. Admin confusion means dependence. Change management becomes a moat.
Where AI Starts to Replace—And Where It Amplifies
AI isn’t here to fire your system admin. But it is here to change who you hire, how they work, and what you expect from them. And that starts by understanding what system admins actually deal with every day.
Here’s a breakdown of how AI can meaningfully assist—or outright absorb—core admin functions:
1. Configuration & Release Note Interpretation
Current State:
Admins parse dense release notes to identify how compliance changes, bug fixes, or interface updates impact their unique setup. One small change to a field rule can create a domino effect across templates, user roles, or disclosure logic.
AI Assist:
We’re working on an AI assistant that can ingest each new release’s notes AND your LOS configuration to:
- Summarize what changed and why it matters based on your configuration
- Suggest test cases for validation
- Flag dependencies (e.g., “This affects Field ID 3647, used in your Rate Lock form”)
Upshot:
Saves hours per release cycle per admin, reduces risk of compliance gaps, and results in less downtime from unexpected/unhandled changes.
2. User Access & Role Management
Current State:
Admins manually assign roles, adjust permissions, and manage user provisioning—often with poor visibility into who needs what or why access breaks. The complexity of the LOS ecosystem mandates complex configurations, that part is unavoidable. What a good SA provides is consistency in what results.
AI Assist / Replace:
With smart AI tooling:
- New users can request access via natural language (“I need access to pricing and disclosures”) and AI can make sure to vet all of the relevant edtails, such as business purpose, limits, etc
- At the simplest level, AI can map-out all of the work to be done and queue it up for your SAs to execute. As AI advances, no SA will be required for rote tasks resulting in the only bottleneck being approvals.
- AI can proactively flag when users have excessive or unused permissions, improving security posture
Upshot:
The majority of the access management burden is offloaded, improving agility and compliance, and leaving SAs to focus on tasks that require more brainpower than key-tapping.
3. Field Logic, Business Rules, and Custom Validations
Current State:
Admins build and troubleshoot validations, calculations, and workflows. One misplaced condition can cause loan data sync failures or misprice a loan. Or, more likely, one gap that allows users to circumvent a rule can result in a buyback.
AI Assist / Usurp:
AI can:
- Write and test field logic based on prompts (“Create a trigger to show Section C fees only when loan type is FHA and property is in Texas”)
- Audit existing rules for conflicts or inefficiencies or gaps
- Auto-generate regression tests to validate changes
Upshot:
Fewer rule errors, faster testing cycles, and the ability for non-admins to self-serve logic suggestions.
4. Bug Triage & Support Ticketing
Current State:
Admins spend hours diagnosing problems (“Why isn’t the appraisal fee populating?”) that are often caused by upstream data, vendor integrations, or broken logic.
AI Assist:
- AI pre-analyzes logs, compares user behavior, and provides likely root causes
- Suggests past workarounds based on community data or internal notes
- Can even open a draft support ticket with complete details for your LOS vendor
Upshot:
Faster resolution, less frustration, and a meaningful reduction in escalated tickets.
Why It Matters Now
System admins aren’t being replaced—but the definition of “qualified” is changing fast.
You still need humans to make judgment calls, handle exceptions, and design elegant workflows. But the clerical, repetitive, high-cognitive-load work? That’s exactly where AI thrives.
If you’re still scaling your ops by throwing headcount at complexity, ask yourself:
- What if your AI tools could eliminate 80% of the repetitive tasks your admins deal with?
- What if they could give your team the clarity to move faster, plan smarter, and execute cleaner?
- What if you didn’t need to hire three more admins just to keep pace—but could empower one with the tools of ten?
This isn’t an experiment. It’s not a pilot. It’s the new minimum standard for competitive ops.
What Lenders Should Be Doing Today
1. Audit Your Admin Workload
Break down your system admin time into categories: user access, rule creation, release review, integration support, issue triage. You’ll likely find that more than half of the effort is reactive and repetitive—prime candidates for automation.
2. Train AI on Your Environment
Start building an internal GPT or AI assistant. Feed it your admin guides, release notes, field list, custom logic. Or have a vendor like Brimma build one for you. You don’t need to boil the ocean—begin with high-impact areas like disclosure logic, pricing issues, or permission setup.
3. Rethink Hiring Criteria
Stop looking for LOS wizards. Start looking for automation architects. Your next best admin may not be someone who knows where every setting lives—but someone who knows how to reduce settings that need to be touched in the first place.
4. Lead From the Front
If you’re in a leadership role, model the mindset shift: that AI is not a nice-to-have—it’s a multiplier. Signal that automation isn’t about job cuts. It’s about elevating every role to focus on work that actually matters.
Final Thought: AI Isn’t Replacing Admins—It’s Replacing Excuses
The best admins won’t be replaced. They’ll be elevated. But the ones who rely on tribal knowledge, opaque processes, or “because that’s how we’ve always done it” thinking? Their days are numbered.
The future of system administration isn’t reactive. It’s predictive, proactive, and AI-assisted.
And if you’re not building toward that future—you might already be falling behind.
Visit our website to learn more!
