Addy AI
Addy AI builds custom AI agents for US mortgage lending: loan officers and processors get automated document review, condition clearing and borrower follow-up inside their existing LOS, POS and CRM. The vendor claims loans close 90% faster.
What is Addy AI?
Addy AI is a set of AI agents built for one industry: mortgage lending in the United States. The publisher is Addy AI, Inc, a US company registered in Michigan, whose terms of service give a head office at 2601 Cambridge Ct, Suite 135, Auburn Hills, MI 48326, while the privacy notice uses a San Francisco mailing address at 945 Market Street, Suite 501. The domain addy-ai.com was registered with NameCheap on 29 December 2022 and first archived a day later; the site has since moved to addy.com, though the older domain still appears in both legal documents. The starting point stated on the About page is blunt: a mortgage takes an average of 45 days to close, with weeks spent reviewing applications, verifying documents and chasing paperwork. Addy AI answers with agents trained on the lender's own data, guidelines and rate sheets. The company claims loans close 90% faster and that a clear-to-close-ready file can be submitted in five minutes. The product breaks into five bricks: AI Mortgage Lending, an AI Loan Officer Assistant, Document AI, an AI email client and a browser extension. Document AI combines computer vision, OCR, deep learning and human validation to pull structured data out of 1003, 1040, 1099, W2 and W9 forms, bank statements and pay stubs. A guidelines chatbot answers questions in natural language and cites its source, whether Fannie Mae, Freddie Mac or a non-QM lender. The assistant will also phone borrowers to ask for missing documents. The deployment posture is the distinctive part. Addy AI runs inside systems the lender has already cleared: Encompass (ICE Mortgage Technology), MeridianLink, Salesforce, HubSpot, Bonzo, Dark Matter, Go High Level, Microsoft Teams and Slack. The vendor puts it plainly: We don't ask lenders to move their data to us. Leadership mixes industry and tech: Michael Vandi (CEO, ex-AWS, Carnegie Mellon), Jessica Rosillo (COO, ex-United Wholesale Mortgage), Shaun Groves (CRO, ex-UWM) and Joao Melo (Head of Engineering, ex-Deloitte). Named customer testimonials, trade press coverage in HousingWire and National Mortgage Professional, and claimed figures such as 2,699 hours saved per loan officer per year support the pitch. The scope, however, stays strictly American and strictly mortgage.
What it does
- Check a loan file against agency or lender guidelines and flag the documents that are missing
- Extract structured data from 1003, 1040, 1099, W2 and W9 forms, bank statements and pay stubs
- Clear loan conditions and AUS findings against the documents already in the file
- Chase borrowers and brokers by email and by voice call for missing paperwork
- Answer guideline questions in plain language with the source cited
- Sync loan data with the LOS, the POS and the CRM
- Sort, summarize and draft the loan officer's email, and spot past clients worth a refinance call
When to use Addy AI / When not to
A quick filter to help you decide if Addy AI is the right fit.
When to use Addy AI
- US mortgage lenders and banks handling a steady volume of loan files
- Loan officers buried under document collection and borrower chasing
- Processing teams already running Encompass, MeridianLink or Dark Matter that want AI grafted onto the LOS they own
- Lending shops that want to absorb more files without hiring: one named customer reports that a single employee now handles four times more loans
- Organizations with enterprise security requirements, given the SOC 2 Type 2 attestation and GLBA-aligned controls
When not to use Addy AI
- Anything outside mortgage: the agents are trained on Fannie Mae, Freddie Mac and non-QM guidelines, not on general business documents
- Lenders operating outside the United States: the compliance framing is GLBA, CCPA, TCPA and RESPA, the governing law is Californian, and there is no EU infrastructure or representative
- Buyers who need a published price and a self-serve signup, since there is no pricing page, no free trial and no free plan, only a sales demo
- Teams that want a public API to build against, because none is documented
- Individual borrowers: the product is sold to the lender, not to the person taking out the loan
How to use Addy AI
A typical end-to-end flow, from setup to results.
- Book a demo through the request-demo page, or use the Calendly and Google Calendar links on the contact page, since there is no self-serve signup for the mortgage offering
- Scope the deployment with the sales team: which product bricks, which lending workflows, which systems
- Complete onboarding, which the terms of service make a condition of the refund window, so expect a structured setup phase
- Train the agents on your own material by uploading guidelines, rate sheets, documents and source links
- Connect the agents to your LOS and POS, then to the CRM (Salesforce, HubSpot, Bonzo, Go High Level) and to Microsoft Teams or Slack
- Set up team access with SSO and SCIM provisioning where IT requires it
- Install the browser extension from the Chrome Web Store to get the sidebar on any site
- Run the AI Loan Processing Checklist on a file and review what the agent flags as missing
- Ask guideline questions in plain language and check the cited source before acting on the answer
- Let the agent chase borrowers by email and voice call, and migrate mailboxes from Outlook, Yahoo or Gmail if you adopt the email client
Pros & Cons
Pros
- Vertical by design: trained on mortgage guidelines rather than sold as a general-purpose assistant
- Installs into systems the lender has already cleared, such as Encompass and MeridianLink, instead of demanding a data migration
- Documented and audited security posture: SOC 2 Type 1 and Type 2 attestations, GLBA-aligned and CCPA-aligned controls, detailed encryption and per-tenant isolation
- A clear commitment not to train on customer data, repeated both in the Trust Center and in the privacy notice
- Guideline answers cite their source, which makes them checkable rather than merely plausible
- Broad coverage of the cycle: documents, conditions, borrower communication, email and a browser extension
- Leadership drawn from the industry itself and from tech, backed by named customer testimonials, a published subprocessor list and SOC 2 reports available under NDA in about one business day
Cons
- No public pricing at all: no pricing page, no amount, no tier, so you cannot position yourself without talking to sales
- Neither a free trial nor a free plan is documented anywhere on the site
- Strictly mortgage and strictly American: guidelines, compliance framing and arbitration are all anchored in the United States
- No documented public API, so integration has to go through the connectors the vendor provides
- No Article 27 GDPR representative, no DPO and no published DPA, despite a privacy notice that acknowledges collecting EU residents' data
- The hosting country is never named, the Trust Center going no further than an enterprise-grade cloud provider
- User counts contradict each other across the vendor's own pages, 16,732 on the Email Client page against 11,972 on the extension page, and the legal documents age unevenly, with terms dated 17/05/2025 against a privacy notice dated 29/06/2026
Pricing & Plans
Addy AI publishes no pricing. There is no pricing page on the site and no pricing URL in the 93 KB sitemap, so no entry price and no currency can be quoted here. No free plan and no free trial is documented anywhere: every call to action leads to Book a Demo, Request Access or a Calendly booking. The terms of service do confirm the commercial model: the service is sold as a subscription with recurring billing that continues without re-authorization until cancellation, All payments shall be in U.S. dollars, and Visa, Mastercard, American Express, Discover and PayPal are accepted through Stripe. Subscribers may cancel at any time from their account or by email, with effect at the end of the paid period. A full refund is available between the 91st and the 120th day of subscription, subject to completed onboarding, 90 consecutive days of use and compliance with the terms, and is paid within five to seven business days through Stripe.
- the site names no tier and no price point
- Only two entry points exist
- Book a Demo for the mortgage offering and Request Access through the app for the email client
- AI Mortgage Lending
- AI Loan Officer Assistant
- Document AI
- Email Client and the Chrome extension
- Positioning is stated as built for teams of all sizes
- with SOC 2 requirements addressed for larger organizations
Data, GDPR & hosting
A consolidated view of how Addy AI handles your data.
GDPR overview
Addy AI nowhere claims to be GDPR compliant, and its Trust Center lists SOC 2, GLBA and CCPA but never the GDPR. The privacy notice does cite the GDPR and the UK GDPR and sets out the legal bases relied on: consent, contract performance, legal obligations, vital interests and legitimate interests. EEA and UK residents are offered access, rectification, erasure, restriction, portability, objection and the right not to be subject to automated decision-making, plus a complaint to their supervisory authority; Swiss residents are pointed to the FDPIC. Consent can be withdrawn at any time, and rights are exercised through a data subject access request form or by emailing info@addy-ai.com. The gaps are real: no Article 27 representative, no DPO, no published DPA, and a declared absence of international transfers that sits awkwardly with a US publisher addressing EU residents.
Who owns the data?
Addy AI, Inc. is the controller. Its privacy notice, last updated 29 June 2026, states that it is responsible for making decisions about how personal information is processed. Collected data covers names, email addresses, sent emails and email threads, with email content itself declared sensitive, plus location, device, log and usage data; Google Workspace users also expose Gmail content, Calendar events, contacts and metadata. Customers keep control of their loan data in practice: the Trust Center claims per-lender logical isolation across storage, queues and model contexts, and the notice states We do not retain this user data for the purpose of developing, improving, or training generalized AI and/or machine learning models. Deletion is available on request, with residual retention for fraud, disputes and legal obligations. No DPO is appointed and info@addy-ai.com is the only contact.
Reuse rights
The published documents frame data use narrowly. The contractual purposes declared are user-to-user communication, delivering the service, sending administrative information, answering requests and managing orders; the legitimate interests declared are protecting the service, requesting feedback and identifying usage trends. Sharing is limited to business transfers, meaning a merger, sale or financing event, and to affiliates, and the CCPA disclosure states that personal information is neither sold nor shared. Loan data, documents and, for voice calls, the audio and transcripts are passed to third-party AI providers after explicit in-app consent; those providers act as our processors under agreements requiring data protection equivalent to ours, and may not use your data for their own purposes. Payments are handled by Stripe. Cookies and tracking technologies are in use, including third-party analytics and advertising, and automated SMS loan-status messages can be stopped by replying STOP.
Data retention & training
Hosting summary
No hosting country and no region is named anywhere on the site. The Trust Center takes the question head-on and answers only this: Where is data hosted? With an enterprise-grade cloud provider. All data is encrypted at rest with AES-256 using built-in key management. There is no provider name, no jurisdiction and no data residency option. What is documented is the protection rather than the location: AES-256 with 256-bit keys across all databases, data stores and file systems, TLS 1.2+ in transit including external transmissions and sensitive email, and per-lender logical isolation across storage, queues and model contexts. A subprocessor list is published in the Trust Center with change notifications, each subprocessor reviewed against security and confidentiality criteria. Two indirect signals exist, neither of which settles the question: the Termly configuration behind the terms of service records a website hosting country of United States, a value never displayed in the rendered document, and the addy-ai.com domain resolves to 199.36.158.100, a Fastly anycast node that says nothing about where data is stored. Anyone with a data residency requirement should ask before signing.
Things to keep in mind
Risks and trade-offs to weigh before adopting Addy AI.
- The terms of service place sole responsibility for TCPA, Telemarketing Sales Rule, CAN-SPAM and RESPA compliance on the customer, for communications the tool itself generates, and the publisher expressly disclaims liability for fines, penalties, regulatory action or loss of licence
- Liability is capped at the lesser of the amounts paid over the previous six months or 100 USD, a token sum against the regulatory exposure of a mortgage operation
- Disputes go to mandatory arbitration in Santa Clara County, California, after 30 days of informal negotiation, with a contractual one-year limitation period
- Loan documents and, for voice calls, the audio and transcripts are passed to third-party AI providers after in-app consent, which is worth understanding before enabling voice follow-up on borrower files
- No Article 27 GDPR representative and no DPO, even though the privacy notice acknowledges collecting data from EU residents, and that same notice declares no international transfers, which is hard to reconcile with a US publisher
- AI output still needs a human check: the publisher states that it does not guarantee regulatory compliance of the communications generated, so a processor who stops reading the file will eventually miss what the model missed
- The legal documents age unevenly, terms last updated 17/05/2025 against a privacy notice dated 29/06/2026, so check which version governs before relying on either
Setup & Integrations
Technical difficulty
Moderate, and heavier than the marketing suggests. The vendor promises to go live in minutes with no complex setup required, and the browser extension really is a one-click install. The enterprise reality is denser: agents must be connected to the LOS, POS and CRM, then trained on the lender's own guidelines, rate sheets and uploaded documents; SSO and SCIM provisioning imply IT involvement for a team rollout; and the terms of service make completed onboarding a condition of the refund window, which points to a structured, assisted deployment. There is no self-serve path for the mortgage offering.
Deployment
Integrations
Supported languages
Behind Addy AI
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
Who is Addy AI for?
How much does Addy AI cost?
Is there a free trial or a free plan?
Is my data used to train AI models?
What certifications does Addy AI hold?
Where is the data hosted?
Which tools does Addy AI integrate with?
Is there an API?
Is there a mobile app?
Can Addy AI be used outside the United States?
Should you pick Addy AI?
Addy AI is a serious vertical tool aimed at one market: mortgage lending in the United States. Its strength is placement. Rather than asking a lender to move loan files into yet another platform, it runs inside Encompass, MeridianLink, Salesforce or Microsoft Teams, systems that have already passed the lender's security review, and it backs that with a documented posture: SOC 2 Type 1 and Type 2 attestations, GLBA-aligned and CCPA-aligned controls, AES-256 encryption at rest, TLS 1.2+ in transit, per-lender logical isolation, and a repeated commitment that customer loan data never trains shared models. Its main weakness is commercial opacity. There is no pricing page, no published tier, no free trial and no free plan; the only way to learn what it costs is to book a demo. For a buyer building a shortlist, that alone is a real obstacle. For a European reader there is a second blind spot. The privacy notice cites the GDPR and the UK GDPR and lists legal bases and data subject rights, but no Article 27 representative is appointed, no DPO is named, no DPA is published, and the hosting country is never disclosed. The terms of service, last updated in May 2025, also place full responsibility for TCPA, CAN-SPAM and RESPA compliance on the customer, with liability capped at 100 USD. The practical verdict is narrow and clear. If you run a US lending operation on an established LOS and your team is drowning in document collection and condition clearing, Addy AI is worth a conversation, and the named testimonials give you people to ask. Anywhere else, outside mortgage, outside the United States, or on a budget that needs a published price, it is the wrong tool.
- Choosing a selection results in a full page refresh.
- Opens in a new window.