Vesta review
Vesta is a Loan Origination product from Vesta Innovations, Inc. (independent, venture backed). MortgageTechReview scores Vesta 4.3 out of 5.0, ranking Vesta #5 of the 32 products tracked in Loan Origination Software, as of August 11, 2026. Scores on MortgageTechReview are weighted across five axes and are never paid for or influenced by a vendor relationship.
AI-native LOS founded in San Francisco in 2020. Backers include Andreessen Horowitz, Bain Capital Ventures and Conversion Capital, with roughly $35 million raised. Built around AI agents that work loans alongside people from application through funding. Selected by Pennymac in September 2025 and by New American Funding in July 2026, with that rollout phased through 2027. The architecture and the results are arguably the best on this list. It ranks fifth for one reason. The deployment evidence sits with a handful of very large lenders, and part of it is still ahead of us.
How Vesta compares to Encompass
Ranked first in LOSEncompass currently scores highest in LOS, so every other product in the category is compared against it here. That is a ranking on our published rubric rather than a recommendation, and it changes when the scores change. Category Leader
| Axis | Vesta | Encompass |
|---|---|---|
| Production impact | 4.6 | 4.9 |
| Functionality & depth | 4.3 | 4.9 |
| Integrations & ecosystem | 4.6 | 4.9 |
| Adoption & support | 4.0 | 4.4 |
| Return on spend | 3.7 | 3.9 |
| Overall | 4.3 | 4.7 |
Vesta wins 0 of 5 axes against Encompass, on the weight profile published for this category. Full head-to-head →
Where it wins
- The best publicly verified outcome metric on this entire list. Pennymac reported cutting its operational cost to originate by 25 percent on its fourth-quarter 2025 earnings call. That is a public company statement to investors, not a vendor case study.
- Genuine agentic execution rather than AI assistance. You assign work to an agent the way you would assign it to a person. It escalates what it cannot resolve. Agents run disclosures, AUS, fee pulls, and third-party orders unattended.
- Document handling splits, classifies and extracts from any document type, with no model to train. That is the specific problem consuming processor hours.
- Full audit trail, including agent reasoning and the artifacts used. That is what makes AI deployable in a regulated workflow.
- Task-based architecture with no-code rules and modern APIs, plus chat-based configuration. Two of the largest IMBs in the country picked it over incumbents. That is the strongest validation the architecture could get.
Where it falls short
- Founded 2020, and the New American Funding rollout does not complete until 2027. The proof is genuinely promising but still partly in the future.
- Small customer base concentrated in a few very large lenders. Little evidence exists on how it performs for a mid-size lender with a small ops team. The platforms above it deploy today across a range of shop sizes.
- The named wins are genuinely impressive but they are two lenders. Two data points, however good, is not yet a track record.
- $35 million raised is modest against incumbents with hundreds of millions behind them. The multi-year vendor risk is real in a way it is not with ICE or Constellation.
- No published pricing, and an AI-agent pricing model is a new enough thing that the cost structure deserves careful reading.
Why it scores 4.3
Scored on the Loan Origination Systems weight profile. The number shows where it sits in this category. It rests on evidence anyone can check, including the vendor's own record. The weights →
Production impact
25% of scoreThis is the strongest evidence on the list. Pennymac told its fourth-quarter 2025 earnings call that operational cost to originate fell 25 percent. That is a public company statement to investors, not a vendor case study. Pennymac has also expanded the platform past a single production channel, a harder test than a one-line pilot. Vesta separately reports a 988 percent rise in AI agent usage over three months. That is an adoption figure, not an outcome figure.
Functionality and depth
20% of scoreThis is a full origination system, not a workflow wrapper. Coverage runs from application through funding, a rules engine lets administrators change process, and task routing sends work to a person or an agent. Agents run disclosures, AUS, fee pulls and third-party orders unattended and escalate what they cannot resolve, with a full audit trail including agent reasoning. The far end of the file is thinner in public documentation: servicing handoff, post-close audit and channel-specific requirements.
Integrations and ecosystem
25% of scoreVesta claims over one hundred prebuilt integrations, and the public catalog backs a large share with real vendor names. Credit alone lists sixteen bureaus and resellers, among them Xactus, Factual Data, Informative Research and MeridianLink. Pricing covers Optimal Blue, Polly, Lender Price, LoanPASS and Mortech. Both agency AUS engines are connected. Verification runs from Truv and Argyle through to Equifax The Work Number. The catalog thins at the edges: servicing shows one partner, closing shows Snapdocs alone.
Adoption and support
20% of scoreOperations leaders can change rules without filing a vendor ticket, which answers the most common complaint about legacy origination software. Two of the largest IMBs in the country picked it over incumbents. The evidence stops at the implementation itself. There is no public detail on conversion timelines or professional services cost, support staffing goes undescribed, third-party review volume is close to zero, and the New American Funding rollout runs into 2027. Ask for a reference on a completed migration.
Return on spend
10% of scoreThe 25 percent cut in operations cost at Pennymac is the return argument, and it is a good one. Pennymac is also an enterprise with the scale to redesign process around new software, so a mid-size lender should not assume the same result transfers. Nothing about price is public, and an AI-agent pricing model is new enough that the cost structure deserves careful reading. Running two systems through conversion stretches the payback period further.
On price. Quote only. No tier or minimum appears on the site, and no per-loan figure either. Terms are negotiated per lender. Price the migration and the parallel-run period separately from the license. On an origination system replacement, those two dominate year one spend.
