Dots for realtors

OpenAI introduced Dots on September 29. Powered by GPT-6 Astra, each dot has a cloud computer and can work toward ongoing goals, retain context and use connected applications. OpenAI says its plugin ecosystem connects to more than 4,000 apps. Dots are accessible through ChatGPT, with Slack and Teams also described as communication channels. OpenAI’s Dots announcement

For a realtor, the useful shift is continuity. You could establish how you prepare for a listing appointment once, then ask the assistant to prepare the next appointment using the relevant client information and files. The intended result is work that advances between conversations.

Access is rolling out gradually. OpenAI lists Pro in eligible markets, Business Premium and an administrator-enabled Enterprise beta. A first dot is included with eligible plans, with an allowance for deeper work; this does not mean unlimited execution. Setup begins on desktop. The current help documentation says dots cannot initiate calls to you at launch, and texting is a limited beta for some U.S. Pro users. Canadian agents should not assume SMS access. Dots setup and availability

One distinction matters for ongoing follow-up: Dots’ proactive research uses read-only tools. It can identify something worth acting on, but that research process cannot itself send a client message, change a CRM record or operate a browser. Follow-up actions pass through the normal permissions and action checks. OpenAI’s explanation of proactive research

A useful deployment therefore needs both a clear responsibility and the connections required to carry it out.

The following are workflow designs to test with Dots or another capable agent. They describe useful outcomes and dependencies, rather than claiming that every product delivers them out of the box. Each starter is intentionally simple; the assistant should use saved context and ask for missing details conversationally.

1 Prepare the day around clients and deadlines

A useful morning briefing would reconcile the calendar with recent email and CRM activity. If a buyer meeting is scheduled for 11 a.m., it should retrieve the latest conversation, surface unanswered questions and prepare a short agenda. If two appointments leave too little travel time, it should identify the conflict.

The finished output should be a short list of priorities, appointment briefs and draft responses. “You have twelve unread emails” is much less useful than “Your seller asked whether the price reduction is happening before the weekend; a decision is needed before the marketing update.”

Email and calendar connections provide a starting point. A CRM connection adds the client’s history, stage and next action. Deal deadlines need their own reliable source.

Measure whether preparation gets faster and whether important commitments stop falling through the cracks. A longer briefing is not necessarily a better briefing.

2 Turn a showing conversation into useful follow-up

Imagine a buyer likes the location but dislikes the layout, raises their budget and rules out homes backing onto a busy road. The assistant should turn those observations into structured updates: revised criteria, property-specific feedback, unresolved questions and a draft recap.

With authorized CRM access, it could update the correct buyer record and create the next task. With an appropriate listing connection, it could prepare a revised shortlist. The email should explain why each home fits and identify compromises, rather than sending another undifferentiated batch of listings.

The difficult part is preserving meaning. “Would consider a townhouse” is not the same as “only wants townhouses.” A passing comment about price should not overwrite a confirmed budget without clarification.

Success means the next search and the next conversation reflect what the client actually said.

3 Prepare a listing appointment and defensible CMA

The assistant could organize the seller’s questions, property details, renovation history, photographs and prior correspondence. It could assemble a presentation and draft explanations of competing listings, recent sales and positioning.

For the comparative market analysis itself, the workflow needs relevant, authorized comparables. Each should retain its address or identifier, status, relevant dates and source. The analysis should explain exclusions and adjustments: a renovated detached house on a premium lot may be a poor benchmark for an unrenovated home on a main road.

The finished package should distinguish the evidence, the pricing range it supports and the judgment still required from the realtor. Missing basement information or an uncertain property attribute should remain visible.

A general agent can help analyze a supplied comparable set. Automatic retrieval of the right local sold data is a separate capability that must be demonstrated.

4 Keep a listing launch consistent across every asset

Start with a confirmed property record, approved photographs, brand settings and a launch date. From that shared source, the assistant could prepare a property page, feature sheet, email announcement, social captions, short video script and open-house materials.

The greatest benefit may be change management. If the launch date moves, all scheduled materials should move with it. If a property detail is corrected, that correction should reach every draft before publication.

This workflow spans writing, design, websites, file storage and scheduling, making it a strong candidate for a general agent with broad connections.

Real estate specialization adds factual controls. A system should flag an unsupported “legal basement apartment” claim instead of repeating it across six assets. The acceptance test is a coherent campaign using verified facts, with the approved versions delivered to the correct destinations.

5 Produce a seller update that answers the next question

A weekly seller report could bring together showing activity, feedback, marketing results, relevant competing listings and the previous week’s commitments.

The report should explain what changed and why it matters. For example: showing volume may be stable while feedback repeatedly identifies the same objection. That calls for a different conversation than a listing receiving little attention at all.

Useful outputs include a client-facing report, a draft email and a private preparation note for the realtor. The preparation note can identify questions to discuss without presenting an unapproved recommendation as a decision.

The dependencies are easy to underestimate. Email may contain some feedback; the showing platform may contain the rest. Advertising clicks do not establish buyer intent. The workflow needs to reconcile these sources and state gaps instead of filling them with confident prose.

6 Reactivate leads with relevant reasons to reconnect

A useful assistant would identify people whose previous plans make a timely conversation worthwhile: a buyer who paused until a lease renewal, a seller who wanted to revisit their options in the autumn, or a past client who asked about a particular neighbourhood.

It could prepare a prioritized review queue with the reason for contacting each person, the supporting conversation and a suggested message. With the appropriate permissions, the approved activity could then be recorded in the CRM.

A specialized workflow should preserve contact preferences, respect suppression records and avoid treating every database entry as a prospect ready for a campaign. It also needs to recognize when two records represent the same household.

Measure replies, useful conversations, appointments and corrections required. Producing hundreds of drafts says little about whether the outreach helps the business.

7 Turn market research into a repeatable client service

An assistant could collect the latest releases for the agent’s market, update charts, explain the changes and prepare a newsletter, presentation and social content from the same analysis.

The useful discipline is consistency. Every chart needs a geography, period, unit and source. The writing should distinguish average price from benchmark price, monthly movement from annual change, and a national trend from the agent’s local market.

A recurring workflow should detect a new release, check whether historical values were revised, update the analysis and bring the package back for review. Replacing last month’s numbers without reconsidering the explanation is insufficient.

This is a promising meeting point between broad research and document tools and specialized housing data. Start with one market and one dependable source sequence, then expand.

8 Coordinate showings and transaction administration

A showing workflow could collect the selected properties, confirm the buyer’s availability, calculate travel time and propose a route. Execution then requires access to the relevant booking process, appointment confirmations and a way to handle declines or changes.

A useful itinerary distinguishes requested appointments from confirmed appointments. It also keeps access instructions in the appropriate private destination.

Transaction administration follows the same principle. An assistant could assemble a document checklist, flag missing information and prepare reminders from verified dates. Preparing an offer adds further dependencies: the correct current forms, permitted access, accurate parties and property details, and review of transaction-specific terms.

The acceptance test should be concrete: confirmed appointments in the calendar, the correct file updated, or a complete unsigned draft ready for professional review. A message saying “done” is not evidence that the underlying job finished.

These products overlap, but the decisions they ask a realtor to make differ. Our assessment below separates their documented emphasis from the work a realtor should verify.

The product descriptions are supported by the vendor sources linked below. The suggested tests are our analysis.

Dots and Grok Bot

Grok Bot is a direct peer. Its documentation describes cloud computer use across websites and applications, multiple bots coordinating work, and routines learned by watching a user demonstrate a process. Grok Bot announcement

For a real estate team, that suggests a practical experiment: demonstrate the full process for turning a buyer consultation into CRM notes, follow-up and tasks, then evaluate the next several runs. Check record accuracy, exception handling and how often someone has to rescue the process.

Dots offers an OpenAI-based route to continuing work across applications. Grok’s explicit emphasis on coordinating bots and learning routines makes those capabilities worth comparing directly. We have not run a controlled test establishing that either completes realtor work more reliably.

Homies has also published a realtor workflow as a shared Grok Bot template. That illustrates how a general agent and real estate expertise can be combined. It does not make every connected data source or form available automatically. Homies’ Grok Bot field guide

Dots and Meta Muse

Meta describes Muse as a personal agent with a dedicated virtual computer, browser access and a conversational interface that includes WhatsApp. Meta’s Muse introduction

Its September 29 small-business announcement matters particularly for realtors. Meta lists connections including HighLevel, Canva, QuickBooks, Notion, Shopify and Stripe, alongside Facebook and Instagram business accounts. It also documents custom connectors and availability in the U.S. and Canada. Muse for Small Business

That gives Muse a credible position in social marketing and general business administration. For an agent already using HighLevel and Meta advertising, the potential reduction in setup work deserves a test.

HighLevel support also means it would be inaccurate to dismiss Muse as having no realtor-relevant CRM access. The remaining question is what that connection can actually do: read contacts, update records, handle consent fields, reconcile duplicates, or execute the entire follow-up process.

Dots and Instinct

Instinct presents itself as a personal assistant connected to applications and devices, including email, messaging, screen, audio and location. Its public description emphasizes understanding everyday context, following up on dropped threads and arranging practical tasks. Instinct’s product description

For a realtor moving between appointments, that positioning is attractive. Personal logistics and business administration often collide in the same afternoon.

The public material reviewed is too limited to establish equivalent MLS access, transaction workflows or brokerage controls. It also does not justify a detailed feature ranking against Dots. Instinct’s marketing describes texting and calling; actual account availability should be checked before designing a business process around either.

Our assessment is that Instinct warrants testing for personal coordination, with a separate verification step for professional workflows.

Dots and OpenClaw

OpenClaw documents a broad configurable environment: messaging channels, multiple model providers, browser and other tools, skills, plugins, and scheduled work through cron and heartbeats. It supports self-hosted deployment options. OpenClaw’s documented features

The attraction is flexibility. A technically supported brokerage could choose providers and build custom connections around its processes.

That flexibility creates an operational responsibility. Someone must maintain the setup, monitor failed jobs, update credentials and verify that a changed website or API has not broken the workflow. Hosting an agent yourself also does not imply that every model request or connected service keeps data on that machine.

Dots packages more of the operating environment into a managed product. OpenClaw offers a route to greater customization. The right choice depends partly on whether the brokerage wants to operate software infrastructure.

Dots and Hermes Agent

Nous Research documents Hermes Agent with persistent memory, a system for creating and reusing skills, scheduled tasks, messaging integrations, browser tools, MCP connections and several execution environments. MCP is a standard way for an agent to call external tools. Hermes Agent documentation

For real estate, the interesting application is repeated work: improving a seller-report procedure or retaining the steps for a particular CRM process.

A saved or learned procedure still needs verification. It can preserve an incorrect assumption as readily as a useful convention, and a workflow can become outdated when a provider changes.

Hermes is worth evaluating when a team wants configurable tools and reusable operating knowledge. As with OpenClaw, the comparison should include ongoing maintenance, rather than only the ease of the first successful demonstration.

Breadth is a real advantage. OpenAI’s stated app reach and Meta’s published business connectors address a much wider range of activity than real estate alone. These ecosystems are plausible choices for an assistant that needs to help with research, travel, shopping, presentations, accounting coordination and several businesses. The breadth advantage is documented; superiority on a particular job still requires testing.

Consumer convenience matters. A familiar chat application can reduce the effort required to ask for help. Instinct’s personal-assistant approach and Muse’s conversational distribution are meaningful product choices for someone who spends the day away from a desk.

Custom work may suit a configurable agent. A team running a brokerage, a renovation business and a media operation may prefer one environment with separate tools and processes for each. OpenClaw and Hermes provide documented customization paths.

General agents can build specialized capabilities. We should expect competitors and technically capable users to connect realtor tools and package workflows. A connection or prompt is not a permanent competitive advantage.

Homies needs to earn its place through lower setup effort, dependable execution and useful real estate outcomes. It should not assume that agents will pay more simply because a product uses real estate language.

Authorized property data with clear coverage

A general agent’s web browser does not establish access to the complete data a realtor needs. A convincing property workflow should identify the source, geography, listing status coverage, update frequency and permitted use.

The practical questions are specific. Can it retrieve sold comparables for this market? Does it recognize a relisted property? Can it distinguish current records from stale copies? Does the user’s account have the required access?

An MLS plugin could close part of this gap. It would still need reliable field mapping, account permissions and handling for incomplete or unavailable results. The appropriate claim is “coverage not established by the documentation reviewed,” rather than a blanket assertion that a competitor cannot connect.

Working connections to realtor tools

The relevant toolkit may include Follow Up Boss, BoldTrail, Lofty, HighLevel, a showing service, transaction software and an e-signature provider. This is a list of integration requirements to evaluate, not a claim that every named product supports every service.

Reading a contact record is different from correctly maintaining a pipeline. Creating an appointment request is different from receiving confirmation. Generating a PDF is different from completing a signing workflow.

A useful integration needs to preserve the relationship between the client, property, conversation and transaction throughout the job.

Specialized workflows and exception handling

A realtor workflow needs operating knowledge: what information to collect, what to verify, what can happen next and when a human decision is necessary.

Consider a buyer shortlist. It should retain confirmed preferences, explain trade-offs, avoid sending the same property repeatedly and react appropriately when status changes. A listing launch should propagate corrected facts. An offer workflow should identify missing information without inventing it.

The difficult cases reveal product quality: duplicate contacts, contradictory instructions, stale data, unavailable tools and partially completed tasks. Those cases deserve as much attention as the successful demo.

Persistent records and proof of completion

Conversation memory helps an assistant understand the user. The business also needs a dependable record of what happened.

If an email tool times out, the agent should determine whether the email was sent before trying again. If a showing is declined, it should retain that result and revise the plan. If a draft changes after approval, the process should preserve which version was approved.

These are product requirements for dependable execution. They apply equally to general agents and Homies.

Homies publicly packages specialist agents for CMAs, listings, offers, CRM follow-up, marketing, research and reports, with shared brand context. That is its practical proposition: a realtor starts closer to a recognizable job and finished deliverable. Availability still depends on the relevant account, market and tools. The website labels MLS websites as available in select regions and its included native AI CRM as coming soon. Homies product and availability

The external API needs its own description. Homies’ current public documentation lists Canadian market statistics, account-permitted paperwork and U.S. property search through HomeHarvest. It explicitly says licensed IDX/DDF feeds are not yet connected to the unified gateway and Canadian listing search is not enabled there. Canadian evaluations through that gateway require supplied, authorized sold comparables. Homies API coverage

We therefore should not claim that adding the Homies connector gives any agent complete Canadian MLS access.

The API also distinguishes preparing work from delivering it. Buyer tools prepare matches and unsent email drafts; recurring delivery requires a separately configured persistent runner, mailbox authorization, recipient and schedule. Paperwork access depends on account permissions and applicable publisher terms. Homies API execution requirements

This suggests a useful direction: a realtor’s preferred general assistant could call specialized Homies services where the host supports the connection. That is an integration opportunity, not a claim that a tested, approved Homies plugin already exists for every agent discussed here.

Choose a recurring task that consumes time every week. Give each candidate the same permitted inputs, the same objective and the same definition of completion.

For example, ask it to prepare a seller update using a fixed collection of showing feedback, marketing results and comparable information. Then record:

• How long setup and connection took.

• Whether the facts and client record were correct.

• How many corrections and interventions were needed.

• Whether the deliverable reached the intended tool.

• Whether failures and missing information were reported clearly.

• The total cost of the run, including review time.

Repeat the job with a meaningful change, such as corrected property information or an unavailable source. That second run will often teach you more than the first.

Our recommendation is to begin general agents on research, preparation and administrative work where inputs and outcomes are easy to inspect. Evaluate a specialized real estate system on the complete workflows it claims to support in your market. Technical teams can also test a combination: a general agent for coordination, connected to specialist services for particular jobs.

Dots makes this category more compelling. It also raises the standard for Homies. The value has to show up in fewer missed commitments, better client preparation and more work completed correctly.

Bring Homies one recurring job from your actual business. Ask us to show the connections, the finished result and what still needs your judgment. Explore Homies