How I install agentic AI workflows
I install these workflows for real estate operators. The deliverable is the workflow itself, running on your data and producing analytical work at the standard your team holds itself to.
There's more to senior analytical work than judgment. A large portion of the week goes to the assembly that supports it: pulling data from filings and reports, building the model, reconciling versions, walking reviewers through the analysis. The judgment is the part that needs senior time. The assembly is the part an automated workflow can now produce.
What gets installed
Agentic AI runs multi-step work on your team's actual data and tools. It's not a chatbot you ask questions. It's an automated system that handles the production work behind your team's analytical deliverables.
Each engagement begins by mapping where senior people are spending time on assembly. The shape of the workflow varies by client. Workflows are built against your existing templates, conventions, and reviewer standards, not a vendor's. The workflow produces output in those same formats. Common installations include:
- Monthly close and variance reporting. The monthly cycle — pulling data from the property management system, building variance tables, formatting the monthly deliverable. The workflow handles the production; your team's time stays on the judgment that sits on top.
- Portfolio performance review. The recurring analysis on the portfolio — performance against budget, comp benchmarking, surfacing the properties that need attention.
- Acquisition diligence. The data and modeling work behind evaluating an acquisition — submarket research, comp pulls, pro forma scenarios.
- Capital partner reporting. The recurring package for institutional LPs, co-investors, or capital partners — quarterly performance, capital account statements, fund-level summary.
The workflow handles the production work behind your team's deliverables, so more of their week goes to the judgment those deliverables need.
How engagements work
Engagements run in four phases.
A first conversation, 30 to 45 minutes, to walk through the workflow you're considering and whether the engagement is the right fit. No commitment.
If the fit is right, installation scoping begins. Two to three weeks. The output is a written plan for what gets built and how it integrates with your existing systems. $5,000, credited toward installation if you proceed.
Installation runs six to ten weeks. The workflow gets built against the plan and calibrated through real cycles of the actual deliverable, with your reviewers signing off before the workflow runs day-to-day.
Handover at the end. Your team operates the workflow going forward, with an internal AI champion identified and trained during installation to be the workflow's ongoing owner. The systems, models, and processes built during the engagement are yours. No ongoing dependency on this practice.
How an installed workflow runs
The workflow handles the assembly; your team's time goes to the judgment.
The discipline behind the workflow
The workflows that meet the standard share a few characteristics.
The model is the source of truth. The deliverable is the explanation. The Excel models built during installation are where the numbers live. Every figure in every deliverable cites the cell or the source it came from. A reviewer opens the model alongside the deliverable, clicks any number, and finds the calculation in seconds.
The workflow extends your systems, it doesn't replace them. Yardi keeps doing property accounting. Argus keeps handling valuations. Your Excel templates keep their structure and logic. The workflow pulls data from these systems and runs the analytical work on top — variance tables, comp pulls, partner reporting. Nothing in your existing stack changes.
The output matches your team's format, not a vendor's. The workflow produces output that fits the templates your team already uses — partner reporting, monthly close package, diligence package, portfolio review. Calibration happens during installation, against real samples of your team's current work.
Implementation case study
In 2025, a senior buy-side reviewer engaged the practice to install a diligence workflow for publicly-traded REITs. He acted as both client and reviewer; I worked as sole analyst. The engagement ran six weeks. The first run of the workflow produced a deep diligence analysis on a specific REIT under evaluation. The workflow itself remained installed for him to run on any future REIT requiring the same depth of analysis.
The workflow does three things.
Source capture. Public filings, third-party appraisal data, broker reports, and proprietary research get captured into a structured form the model can draw on. Every figure is tagged back to its source document and page. Any number in any eventual memo can be traced to where it came from in seconds.
Model construction. Every property in the portfolio is modeled individually. Third-party appraisals sit in their original form, with adjusted views alongside in parallel columns and an explicit rationale for every change. Forward projections get built two ways — one anchored to actuals, one to management guidance — with a row at the top comparing them so any drift is visible to a reviewer.
Deliverable assembly. The memo is structured to the reviewer's conventions, calibrated during installation against samples of his existing diligence work. Every claim in the prose cites a specific model cell. A reviewer opens the Excel alongside the memo and audits any number in seconds.
The first deliverable cleared institutional review. The workflow runs on each subsequent REIT without rebuilding.
Background
I built the first version of what runs this practice in 2022, and have spent the four years since bringing its output to the standard the work requires. My background spans finance and engineering. Before ChatGPT's public release, I built re.search, a browser extension that generated an AI summary for any search, an early version of what is now standard in Google search. I tested whether LBO modeling could be automated during an investment-banking internship; the next year, at a hedge fund, I built visualization and performance-reporting workflows for the research, sales, and finance teams. By 2025, the technology had matured enough to produce serious analytical deliverables, provided the right discipline was built into the workflow. This practice is the result.
Next steps
Engagements start with a 30 to 45 minute conversation to walk through the workflow you're considering. Common starting points include monthly close, portfolio review, acquisition diligence, and capital partner reporting.
Reach me at ethan@ethanjfarrar.com.
Note: This practice operates through 2026. I'll be continuing this work in-house at a global equity hedge fund afterward. Installed workflows are built for your team to operate independently, with no ongoing dependency.