How to Apply AI in Affordable Housing: Fundraising, Grant Reporting, and Operations

Affordable housing professionals do not need another prediction about AI taking their jobs.

They need to know what AI can actually do for them on Monday morning.

That is the focus of this solo episode of the Affordable Housing & Real Estate Investing Podcast, hosted by Kent Fai He. Kent draws from his previous experience working in technology and AI, along with his current work in affordable housing, to explain where artificial intelligence can realistically save developers, project managers, asset managers, fundraisers, and housing organizations time.

His starting point is simple:

Stop asking whether AI can replace a job. Start asking which tasks inside that job AI can help complete.

Kent explains that affordable housing may actually be one of the industries best positioned to benefit from AI because so much of the work depends on written rules.

States have rules for tax credits. HUD has forms. Lenders have closing checklists. Funders have reporting requirements. Organizations have underwriting standards and internal processes.

Those rules create administrative work, but they also create something AI can learn from.

The opportunity is not to automate the affordable housing professional.

It is to automate parts of the work that consume that professional's time.

Why Is Affordable Housing One of the Best Industries for AI?

Affordable housing is complicated partly because it has so many rules.

Every state can have its own rules for allocating Low Income Housing Tax Credits. HUD has forms and program requirements. Lenders have closing checklists. Funders have reporting requirements. Organizations have their own underwriting standards and approval processes.

Somebody has to understand those rules and apply them correctly across project after project.

Kent believes that is precisely what makes affordable housing a strong use case for AI.

Modern AI systems can work with written instructions, examples, documents, templates, and repeatable procedures. Affordable housing organizations already have enormous amounts of this material.

Consider a project manager working on a complicated affordable housing development.

They may be chasing documents from five or six lenders, funders, grants, or foundations. They may be waiting on environmental studies, geotechnical reports, and other third party due diligence. Then they may take the same project information and reenter it into six or ten different funder templates.

When a cost changes, somebody updates the spreadsheet.

When a funder asks for a report, somebody repackages the information.

When the board needs a memo, somebody turns the same facts into another document.

Kent's recommendation is straightforward:

“Use AI for the boring stuff first.”

Start with work that has clear inputs, written rules, repeatable steps, and outputs that a human can verify.

Should Affordable Housing Organizations Automate Jobs or Tasks?

Kent's first major concept is:

Concept #1: Stop Thinking About Jobs. Start Thinking About Tasks.

One of the biggest mistakes organizations make is asking:

Can AI replace this person?

Can it replace the analyst?

Can it replace the project manager?

Can it replace the grant writer?

Kent argues that these questions are too broad to be useful.

A job is really a bundle of different activities.

An affordable housing project manager might collect documents, update spreadsheets, prepare reports, write board memos, participate in closing calls, walk construction sites, coordinate consultants, and mentor junior employees.

Those tasks should not all be treated the same.

Some are mostly repeatable steps.

Some require judgment.

Some require human relationships.

Some require somebody to physically be at a property.

Instead of evaluating the job title, Kent recommends going one level deeper.

Ask what the person actually does.

Then evaluate each activity separately.

If you reenter the same project information into multiple funding applications, the rules may be relatively consistent.

If you are sitting in a closing call trying to understand why an agency attorney suddenly sounds nervous, you are relying on experience and human judgment.

Those are fundamentally different types of work.

How Do You Find the Affordable Housing Tasks AI Should Handle?

Kent gives listeners a practical exercise.

Write down everything you actually did at work this week.

Do not write:

“Worked on tax credit application.”

Break it apart.

Maybe you:

• Collected site documents

• Ordered an environmental report

• Updated the unit list

• Emailed the architect about drawings

• Updated sources and uses

• Entered information into an application

• Drafted the narrative section

• Prepared a board memo

• Updated a funder report

Then examine each individual activity.

Ask:

Does this have steps?

Are there rules?

Could I write those rules down?

Those are the activities Kent suggests evaluating first for AI.

This changes AI adoption from a vague technology initiative into a workflow exercise.

You are not trying to “implement AI.”

You are finding specific pieces of work where AI can save time.

What Is Kent Fai He's 10, 80, 10 Framework for AI?

Another important idea in the episode is Kent's 10, 80, 10 framework.

The framework divides AI assisted work into three parts.

First 10%: Humans Define Success

The person needs to explain the goal.

What are you trying to produce?

What does good work look like?

What template needs to be completed?

What rules apply?

What information should be included?

For example, if you are using AI to help prepare a funding application, the human first defines the application requirements and provides the relevant project information.

Middle 80%: AI Does the Repeatable Work

Once the instructions and inputs are established, AI can perform much of the repetitive execution.

It might organize information, draft responses, transfer information into a required format, or create a first draft.

Final 10%: Humans Review and Correct

The human comes back in.

Check the work.

Correct mistakes.

Apply judgment.

Make sure the answer actually responds to what was requested.

Kent compares this to training an employee.

If a new employee produces imperfect work on their first day, you normally do not immediately conclude the employee is useless.

You train them.

AI requires feedback too.

What Are AI Skills and Why Do They Matter for Affordable Housing?

Kent's second major idea is about what he refers to as skills, agents, and folders.

The terminology can sound technical.

His explanation is intentionally simple:

Concept #2: Think of an AI Skill Like a Folder of Instructions.

Inside that folder might be:

• Instructions

• Rules

• Examples

• Templates

• Reference files

• Standards for the output

The AI reads those instructions and performs the task according to the way your organization wants it done.

Kent uses a tax preparation analogy.

Imagine choosing between an extraordinarily intelligent person who has never prepared your type of tax return and an experienced tax preparer who has completed a thousand similar returns.

Kent would choose the experienced preparer.

Why?

Because intelligence alone is not the same as relevant experience and context.

An AI model may be extremely capable, but out of the box it does not know:

• Your underwriting standards

• How your state scores tax credit applications

• What your board expects in a memo

• How a specific funder wants information presented

• Your organization's writing style

• Your internal approval process

Without that information, the model has to guess.

The instruction folder gives it the context it was missing.

Why Does Generic AI Produce Generic Work?

This leads to another core idea:

Concept #3: AI Needs to Be Shown What “Good” Looks Like.

Kent explains why so much AI generated writing and work feels generic.

The user often never provided anything specific.

If you type:

“Write me a funder update.”

What exactly should the AI do?

It does not know the funder's priorities.

It does not know your preferred structure.

It does not know the appropriate length.

It does not know your tone.

It does not know what you always include.

It does not know what you would never say.

Kent puts it this way:

“When AI gives you bad work, it is almost never an AI problem. It is an instructions problem.”

He argues that organizations should train AI similarly to how they train people.

Show examples.

Explain the rules.

Define success.

Correct mistakes.

Save those corrections.

Over time, the AI stops relying as heavily on generic assumptions because the organization has documented how it wants the work performed.

What Is the Difference Between AI Connection and AI Know How?

Kent's third major technical concept is especially useful for organizations evaluating AI tools.

Concept #4: Connection and Know How Are Two Different Problems.

Connection means giving AI access to information.

That could include:

• Google Drive

• Project files

• Reporting software

• Salesforce

• Internal systems

Kent references MCP servers in this context and simplifies the concept as essentially giving AI the ability to connect to the systems where information lives.

But connection alone does not tell AI what to do.

That is the second piece:

Know how.

Know how means the instructions, rules, examples, and processes that explain what the AI should do after it accesses the information.

An affordable housing organization could give AI access to every file it owns and still receive mediocre results.

Why?

Because access to information is different from understanding the organization's process.

Kent summarizes the distinction simply:

Connection gets AI to the information.

Know how teaches AI what to do with it.

How Can AI Preserve Institutional Knowledge in Affordable Housing?

This may be one of the most valuable long term ideas in the episode.

Concept #5: AI Instructions Can Become Organizational Memory.

Affordable housing organizations rely heavily on experienced people.

An experienced project manager may know dozens of small things that are not documented anywhere.

They know what a certain funder cares about.

They know how an agency interprets a requirement.

They remember the mistake the organization made three projects ago.

They know how the investment committee likes its memos structured.

Then that person retires or accepts another job.

A huge amount of organizational knowledge can leave with them.

Kent argues that creating detailed AI instructions can help solve part of this problem.

Every rule that gets documented stays.

Every correction can stay.

Every process improvement can be added.

Six months later, the organization has something it may never have had before:

A written record of how it actually works.

That has value even if the organization eventually stops using AI.

Documenting institutional knowledge was always a good idea.

The difference is that AI can create an immediate payoff for doing the documentation.

How Can AI Skills Scale Across an Affordable Housing Organization?

Concept #6: Build a Library of Specialized Knowledge Instead of One Giant AI Prompt.

Kent explains that an AI system does not necessarily need to read every instruction and every document every time somebody asks it to do something.

Instead, different sets of instructions can be organized around different tasks.

Think about labels on filing cabinet drawers.

One might contain instructions for a funder's monthly report.

Another might contain your tax credit application process.

Another could explain board memos.

Another could contain underwriting standards.

Another might explain how the organization handles a particular closing process.

When the task matches the label, the relevant instructions can be pulled in.

This means an affordable housing organization can gradually build a library of specialized processes without creating one enormous, unmanageable set of instructions.

The right knowledge appears when it is needed.

What Parts of Affordable Housing Work Should AI Not Replace?

Kent deliberately saves one of the most important ideas for the human side of the conversation.

Concept #7: Automate Process, Preserve Judgment.

Imagine sitting on a closing call and hearing a change in an attorney's voice that tells you something is wrong.

Imagine walking a property and getting the feeling that a seller may not close.

Imagine managing a difficult lender relationship.

Or negotiating with a partner you have worked with for years.

Those situations depend on experience, relationships, intuition, and judgment.

Kent does not argue that AI replaces those abilities.

The goal is to remove administrative work so people have more time for the parts of affordable housing that need people.

AI cannot solve the housing crisis by itself.

It cannot create missing subsidy.

It cannot manufacture political support.

It cannot replace the relationships needed to move difficult projects forward.

But if an affordable housing professional can save another five or ten hours each week by reducing administrative work, those hours can be redirected toward actually getting housing built.

The 7 Big AI Ideas and Concepts From This Episode

If you remember nothing else from Kent's episode, these are the core concepts:

1. Think tasks, not jobs. Do not ask whether AI can replace an affordable housing professional. Identify individual tasks inside the person's workflow.

2. Use the 10, 80, 10 framework. Humans define success, AI handles much of the repeatable execution, and humans review and correct the final result.

3. Treat AI skills like folders of instructions. Give AI the rules, examples, templates, and standards it needs to perform a specific task your way.

4. Show AI what good looks like. Generic instructions create generic results. Examples and feedback improve the system.

5. Separate connection from know how. Access to your files or software does not automatically teach AI what to do with the information.

6. Turn tribal knowledge into institutional knowledge. Document processes and corrections so knowledge does not disappear when an experienced employee leaves.

7. Automate process, preserve judgment. Use AI for repeatable administrative work while keeping relationships, intuition, negotiation, mentoring, and important decisions human.

Practical Affordable Housing Tasks Kent Says to Look At

Based on the workflows Kent discusses, affordable housing organizations can start by evaluating repetitive work involving:

• Funding and grant application templates

• LIHTC application processes

• Lender and funder checklists

• Board and investment committee memo first drafts

• Recurring funder reports

• Project information entered across multiple templates

• Spreadsheet updates that follow defined rules

• Document collection and organization

• Internal reporting formats

• Organization specific underwriting processes

The important word is evaluate.

A task involving financial, compliance, legal, tax credit, or underwriting information still needs appropriate human review.

AI can help perform the process.

Accountability remains with the people running the project.

Best Quotes From Kent Fai He

“I need you to stop thinking about jobs and start thinking about tasks.”

“Use AI for the boring stuff first.”

“I just wrote down the rules.”

“When AI gives you bad work, it is almost never an AI problem. It is an instructions problem.”

“You have to show them what good looks like.”

Key Insights and Frameworks

• Affordable housing's complexity creates an AI opportunity. The industry contains large amounts of rules, forms, checklists, templates, reporting requirements, and repeatable administrative work.

• Task mapping should come before automation. Break a role into individual activities and identify which ones follow documented steps.

• Instructions create better AI results. Provide examples, rules, tone, structure, required information, and definitions of success.

• AI can help preserve institutional knowledge. Processes that once lived primarily in employees' heads can become reusable organizational instructions.

• Human expertise becomes more valuable, not less important. The highest value work still involves judgment, relationships, negotiation, experience, and accountability.

Common Questions This Episode Answers

How should affordable housing developers start using AI?

Write down everything you do during a normal week and break large responsibilities into individual tasks. Look first for tasks with repeatable steps and written rules, then test AI on those workflows with human review.

What is the 10, 80, 10 AI framework?

Kent describes a workflow where the first 10 percent is defining success and giving AI the right inputs, the middle 80 percent is AI performing much of the work, and the final 10 percent is human review, editing, and judgment.

What is an AI skill?

Kent explains the concept using a folder analogy. The folder contains instructions, examples, rules, and other context that teaches an AI system how to perform a particular task according to your organization's standards.

Why does AI give generic answers?

Generic prompts give the system very little information about what success actually looks like. Kent recommends providing examples, structure, rules, preferences, and corrections rather than expecting AI to infer everything.

Can AI preserve knowledge when employees leave?

It can help. By documenting rules, processes, examples, and corrections, organizations create a reusable written record of how work is performed instead of leaving all of that knowledge inside one employee's head.

What's Coming Next?

This episode intentionally stays at the ideas and concepts level.

Kent's goal is to change how affordable housing professionals think about AI before asking them to build complicated workflows.

The next episode moves from theory toward implementation.

Kent previews walking through three things affordable housing professionals can build from scratch, including an example involving tax credit applications and California's tax credit process.

The larger goal remains simple.

If technology can eliminate five or ten hours of repetitive administrative work each week, affordable housing professionals can put those hours back into the work that matters most.

Building projects.

Solving financing problems.

Working with communities.

Building relationships.

And ultimately creating more affordable housing.

Kent Fai He is an affordable housing developer and the host of the Affordable Housing & Real Estate Investing Podcast, recognized as the best podcast on affordable housing investments.

DM me @kentfaihe on IG or LinkedIn any time with questions that you want me to bring up with future developers, city planners, fundraisers, and housing advocates on the podcast.

Disclaimer: This content is for informational and entertainment purposes only. It is not legal, financial, investment, insurance, or tax advice. It is not an offer or solicitation for any investments. Always do your own research before making investment decisions.