Mortgage AI works best when it supports a defined service process rather than substitutes for one. In his appearance on The Loan Officer Podcast with Dustin Owen, Abdel Khawatmi describes using AI, documented workflows and trained administrators to prepare borrower presentations, support referral partners and reduce repetitive work. The central lesson is practical: document how you want the business to operate, give each task an owner, and measure efficiency without losing the communication that earns client trust.
Conversation originally published January 29, 2026. Recap added October 24, 2026.
Watch the original conversation on YouTube.
Key Takeaways: Automate Preparation, Protect the Relationship
Abdel’s approach centers on three operating categories: sales, service and support. That framework makes automation decisions more specific. Instead of asking where AI could fit, a branch can ask which sales preparation, client communication or administrative task repeatedly consumes time and whether its output can be standardized.
Several practices stand out: prepare cost comparisons early, document procedures before delegating, preserve the original source of each lead, and use administrators to turn an originator’s expertise into repeatable materials. Technology is not presented as a replacement for the borrower conversation. It helps the team arrive at that conversation prepared.
He identifies Monday, Wednesday and Friday updates on post-contract files as the primary practice he would implement if starting his career over. His strongest management question is equally direct: what does it cost to complete a loan without weakening the client experience? Together, those priorities keep automation connected to both service quality and operating discipline.
The Business Started With a Communication Gap, Not an AI Tool
Before mortgage origination, Abdel worked in a personal injury law firm, including a settlement management role. He describes building procedures, monitoring deadlines and supporting negotiations. He was not practicing as an attorney; the transferable experience was organizing complex work including standard operating procedures and monitoring deadlines.
He also saw a recurring problem when clients moved toward homeownership: they complained that their loan officer had not called them back. That service gap helped shape his move into mortgages. He became licensed in 2018 and went full-time in late 2019 after developing business through relationships and purchase transactions.
This background explains why his technology discussion keeps returning to communication. A faster internal process matters only if it helps someone understand their financing, receive a timely response or move through a transaction with fewer unanswered questions. The workflow starts with a service promise, not a software subscription.
For an originator evaluating automation, the practical starting point is therefore a list of recurring communication failures. Where do clients repeat themselves? Which handoffs leave an agent waiting? Which questions require the owner to interrupt another appointment? Those are more useful design inputs than a catalog of AI features.
Bring Payment Expectations Into the Conversation Early
Abdel describes a purposeful five-step intake process, with particular emphasis on its opening stages. First, the team learns what the buyer wants: location, bedrooms, bathrooms and other property requirements. A subsequent conversation, preferably on Zoom, explores the buyer’s hopes and goals with financing comparisons already prepared.
His team uses a total cost analysis, or TCA, and purchase comparisons to make financing more understandable. He describes using ChatGPT to generate specific prompts based on applications and then using Gamma to curate the PowerPoint presentations. Administrators prepare the materials so the originator can spend the appointment interpreting choices rather than assembling slides.
The underlying distinction is important: a loan amount someone may qualify for is not necessarily a payment they want to carry. Introducing estimated costs early can surface that mismatch before a buyer and agent spend substantial time pursuing the wrong price range. The value is the timing and clarity of the discussion, not merely the appearance of the presentation.
A responsible implementation needs an explicit review step. Calculations, assumptions and financing descriptions should be checked by an authorized mortgage professional before delivery. Borrower information belongs only in company-approved systems, and recorded calls or shared summaries require appropriate permissions and controls. AI can help explain a comparison; it should not become the authority on qualification or affordability.
Measure Cost per Loan Without Confusing Cuts With Improvement
When Dustin asks what changed the business, Abdel points to technology, automation and reduced staffing rather than a new selling technique or product. He describes reducing his virtual assistant team from seven to three while training the remaining staff to handle substantial responsibilities with AI assistance. His account is an operating example, not proof that another branch should make the same staffing decisions.
More revealing than the headcount discussion is his preferred measure: the cost of getting a loan done while maintaining the client experience. He is interested in whether the operation becomes more efficient, not simply whether its production target becomes larger. Volume and operating health are related, but they are not interchangeable.
For a branch applying that idea, a useful review would separate software spending, administrative effort, originator preparation time and rework. A presentation assembled quickly is not a true saving if a manager must rebuild it. Similarly, fewer employees may create hidden costs if clients wait longer or producers absorb administrative tasks.
Service measures should sit beside cost measures. Track whether promised updates occur, whether presentations need correction and whether the right person handles exceptions. This gives management a more balanced basis for deciding which work to automate, which to delegate and which requires experienced human judgment.
Track the First Source, Not Just the Last Person Who Referred
Abdel describes tracking business across agents, financial advisers, past clients, social media and personal relationships. The useful detail is how he handles attribution. If an agent contacts him after seeing a video about a specific financing scenario, he wants the initial source recorded as social media rather than allowing the agent relationship to obscure what prompted the inquiry.
He names Follow Up Boss as the CRM used in this process and describes Zapier helping capture lead-source information. He then reviews reports at different intervals to understand what is generating business and where additional spending may make sense. The automation supports reporting discipline rather than replacing it.
A practical extension is to retain both pieces of information: the original discovery source and the current referring relationship. They answer different questions. One helps evaluate content and advertising; the other helps manage partner communication. Combining them into one ambiguous field makes both decisions harder.
Consistent definitions matter more than elaborate dashboards. Before connecting applications, decide how staff should classify an advertisement response, an agent inquiry prompted by a video and a past-client introduction. Then carry those definitions through to applications and funded loans. Otherwise, automated reporting simply delivers inconsistent data faster.
Turn Expertise Into Playbooks Administrators Can Actually Use
Asked for an AI starter kit, Abdel begins with ChatGPT and playbook development. His reasoning is straightforward: experienced originators hold large amounts of useful knowledge in their heads, while new hires often receive tools without clear instructions about what to say or do. AI can help organize that knowledge into an initial working document.
He describes separate Got Mortgages resources for consumer education, sales-professional materials and more detailed operational training. Their contents include scripts, templates, visuals, objection handling and onboarding instructions. The distinction matters because a consumer explanation and an employee procedure serve different audiences and should not be treated as the same document.
For software procedures, he describes using Scribe to create screenshot-based instructions. The goal is not merely to say which platform to use, but to show where to click and what comes next. His lead administrators maintain procedures as tools and workflows change and help run onboarding for new originators.
A useful playbook entry should specify the trigger, required inputs, responsible person, expected output and escalation point. Begin with one frequently repeated task, let someone other than its author follow the instructions, and revise the confusing steps. That test turns documentation from a reference folder into an operating resource.
Use a Focused AI Stack With Clear Review Responsibilities
The episode’s starter stack is deliberately functional. ChatGPT helps structure ideas and draft playbooks. Gamma organizes material into presentations. Captions supports video production and short-form repurposing. Abdel’s broader point is to improve a small set of useful workflows rather than continually start new projects across more applications.
He also describes a custom GPT populated with mortgage guidelines, product information and calculation resources to help staff research questions before escalating them to a manager. This is a different use case from drafting marketing copy: an inaccurate answer could affect how an originator understands a borrower’s options.
The practical rule is to treat that assistant as a research aid, not an underwriting authority. Staff should verify answers against the relevant current source and company requirements. A knowledge library also needs an owner responsible for updates; uploading a document once does not ensure that its contents remain applicable.
Dustin frames AI as an assistant that needs leadership, context and training. That is a useful operating model. Give it approved examples, define the intended audience and review its work. Instructions about voice can make a draft sound familiar, but they do not establish the accuracy of its mortgage content.
Make Partner Education a Repeatable Service
Abdel describes weekly coaching for ten real estate teams, organized into two sessions per weekday. The session structure follows a 60-minute framework: 15 minutes each for sales, service, support, and questions. He seeks input from team leaders about the issues their agents need to address rather than delivering the same mortgage presentation regardless of the audience.
Topics include keeping CRM notes current, improving scripts and helping buyers understand unfamiliar steps. Administrators prepare materials from his direction, use AI to develop them, and return drafts for revisions. That division of labor allows the originator’s expertise to guide the session without requiring him to produce every slide.
The model is useful because it makes partner support scheduled and concrete. An originator considering a smaller version could begin with one team, one recurring meeting and one agreed operational problem. Follow-up should establish whether agents used the material and which questions remain unresolved.
His content advice follows the same practical logic. He cautions against delaying publication while waiting for perfect production quality. Useful explanations can introduce an originator’s problem-solving approach before a partner ever calls. The publishing standard should still include accurate claims and required review; the avoidable delay is cosmetic perfection, not substantive checking.
Build Coverage and Culture Around the System
The conversation also shows why automation needs a human operating structure. Abdel describes routing calls and texts to office coverage when he is unavailable. He uses scheduled availability and a virtual office so originators can obtain help and leads can reach someone able to respond.
This is more than a personal time-management preference. A business remains dependent on its owner if every question, inquiry and exception must wait for that person. Effective coverage requires staff to know what they can resolve, what they must document and when to escalate.
His team example is equally specific: a newer originator helps a sales manager with social media while the manager brings deeper product and guideline knowledge. That exchange challenges the assumption that expertise flows only from senior to junior employees. A shared playbook becomes stronger when different specialists contribute.
The practical lesson is to pair standardization with access to people. Procedures handle repeatable work; colleagues help with exceptions and learning. Neither a software stack nor a document library can replace that combination.
Implementation Checklist: Start With One Complete Workflow
Choose one workflow with a visible service outcome. Early borrower comparison preparation, post-contract updates or partner-training preparation are all grounded in this conversation. Write down the current steps, waiting points, handoffs and recurring mistakes before selecting an automation.
Define ownership and safeguards. Identify who prepares the work, who checks it, who communicates it and who handles exceptions. Confirm approved data handling and recording practices before incorporating client information or call summaries into the workflow.
Create a usable SOP. Capture the instructions and screenshots, provide an example of an acceptable result, and ask another team member to complete the process without informal coaching. Revise the instructions wherever they get stuck.
Pilot and measure. Compare preparation effort, rework, completion consistency and client-facing outcomes. Include the time spent reviewing AI output. Keep the service promise intact while determining whether the change produces a genuine operational improvement.
Expand only after the process is dependable. Abdel’s advice is to keep refining while staying rooted in the underlying system. The goal is not a larger collection of tools. It is a mortgage operation that communicates consistently, develops its people and understands the cost of delivering its service.
Frequently asked questions
What should a loan officer automate first?
Start with one repeatable preparation or support task whose output can be reviewed, such as drafting a partner-training outline or documenting a client-update procedure. Map the task and assign its owner first. Avoid starting with an autonomous system that interprets eligibility or sends unchecked financing advice.
How does Abdel use ChatGPT and Gamma together?
He describes using ChatGPT to organize ideas and develop material, then Gamma to turn that material into presentations. Administrators prepare drafts and he requests revisions. The workflow supports borrower conversations and real estate team education rather than replacing the presenter.
Why prepare a total cost analysis early in the mortgage process?
An early comparison helps distinguish the financing amount a buyer may qualify for from the payment they are comfortable carrying. It gives the buyer, originator and agent a clearer starting point for discussing the search. Estimates and assumptions still need professional review and explanation.
How often does Abdel update clients on post-contract files?
He identifies Monday, Wednesday and Friday updates as a practice he would repeat if starting over. A team adopting that cadence should assign responsibility, document each update and communicate urgent developments when they arise rather than waiting for the next scheduled day.
How should an agent referral prompted by social media be tracked?
Abdel records social media as the initial source when an agent reaches out because of content. A useful implementation preserves both that discovery source and the agent’s identity in separate fields, allowing the team to evaluate marketing while maintaining the referral relationship.
What work do Abdel’s administrators handle with AI?
He describes administrators preparing total cost analyses, custom presentations and real estate coaching materials. Lead administrators also create and maintain procedural instructions and support originator onboarding. Those responsibilities rely on documented processes and review, not just access to AI tools.
Can a custom GPT answer mortgage guideline questions reliably?
It can assist with research, but its answer should not be treated as a final eligibility or underwriting decision. Verify the relevant current guideline, lender requirements and product profile. Assign someone to maintain the knowledge library and escalate ambiguous cases to the appropriate mortgage professional.
How can a loan officer begin offering structured real estate team education?
Begin with one team and ask its leader for a recurring operational problem. Build a session around sales, service, support and questions, following the structure Abdel describes. Prepare reviewed materials, agree on a practical next step and use the next session to assess what agents actually implemented.
To talk with PRMG Got Mortgages, call 973-536-0276, visit GotMortgages.com, contact the team or build your mortgage plan.
Explore more Got Mortgages Insights and Abdel Khawatmi’s profile. Explore more conversations at the Got Mortgages Podcast website.
Author: Abdel Khawatmi with PRMG Got Mortgages | 973-536-0276
This article is general mortgage information only. It is not a commitment to lend, a rate quote, or legal or tax advice. Loan approval, terms and pricing depend on a full application, credit review and property details, and market conditions change. Talk with a licensed loan officer and your own legal or tax professional about your situation. NMLS #1712023 | PRMG NMLS #75243.