A Decacorn in Two Years: A Deep Dive into the Strategy of 'Sierra,' the Synonym for US AI Startups

Introduction

What impression comes to mind when you hear 'AI for call centers'? Many might perceive it as a red ocean, crowded with existing players like Zendesk, Salesforce, Genesys, and Intercom.

However, there is a startup that, since its public launch in February 2024, reached $100M ARR in 7 quarters and over $150M by the end of its second year, and as of September 2025, achieved a valuation of $10B (approx. 1.5 trillion yen) (Source: Sierra Blog "Year Two").

It is Sierra, founded by Bret Taylor (former Salesforce co-CEO, current OpenAI Board Chair) and Clay Bavor (formerly of Google Labs).

Why has it been able to achieve such rapid growth in what appears to be a 'red ocean'?

It lies in a three-tiered strategy: 'Enter through the SoA (System of Action), capture the channel interface, and build a proprietary SoR (System of Record, database).' And this is considered to be a significant strategic approach for the application layer in the AI era.

The structure of this article is as follows.

Part 1: What is Sierra?

Sierra provides a platform for building and operating 'corporate AI agents' that bear the company's brand.

Sierra's clients include enterprises, including regulated industries and long-standing giants, such as Cigna (a major health insurer with $240B in revenue), SoFi (a major US fintech), Ramp (a fast-growing B2B expense management firm), Rocket Mortgage (a major US mortgage lender), and ADT (a major home security company founded in 1874). After building a track record with brand-embodying agents like SiriusXM's 'Harmony,' Sonos's 'Sonos AI Agent,' and WeightWatchers' member agent,, adoption has spread across industries including fintech, healthcare, commerce, media, and mobility, with clients such as Discord, Brex, Rivian, Wayfair, Deliveroo, DIRECTV, Tubi, Bissell, and Vans.

What exactly can it do?

• Operates across major channels such as phone, chat, WhatsApp, email, SMS, and ChatGPT

• Not only answers customer inquiries, but also directly accesses backends such as order management systems, CRMs, and billing systems to resolve issues and execute necessary actions (processing returns, changing subscriptions, renewing contracts).

• Learns the company's brand guidelines, past interaction history, and policy documents to converse in the brand's 'voice'.

In short, it goes far beyond traditional chatbots (which answer questions) to become 'autonomous software that solves problems and completes tasks.'

Key facts listed:

Part 2: Not attempting to disrupt existing SoRs—Sierra's architectural philosophy

2-1. The common wisdom of the SaaS era: 'Integrate data and go after the SoR'

The common wisdom of the enterprise software industry over the past 30 years is clearly articulated by Bret Taylor himself on the Cheeky Pint podcast.

Until now, the database = System of Record of each department has been the center of gravity in their respective business domains, and software companies have built their own applications around them.

The typical pattern for new entrants to beat existing players was to build a better integrated database and replace the existing SoR. This is why 'ERP implementation = multi-year IT project' and 'Salesforce implementation = company-wide sales data migration' were the norm.

2-2. Sierra's contrarian approach: 'Leveraging the fact that SoRs are fragmented as a feature'

Sierra consciously avoided this standard path.

There is an episode in Bret's interview that encapsulates the essence of their strategy. A client who had gone through three acquisitions had three CRMs, three ID systems, and three billing systems. The answer in the SaaS era is obvious: 'Let's start with a data integration project.'

However, Sierra's proposal was this:

In other words, instead of integrating existing SoRs, they place an agent on top of them to make decisions through reasoning while referencing multiple SoRs. What Sierra adopted is a 'reasoning-first, bypass existing SoRs' approach.

Why is this strategically critical? There are three points.

Point 1: The decision-maker for implementation is the business department

Database integration projects always require approval from the IT department (CIO/CTO) and incorporation into the company-wide IT roadmap. For Japanese companies, this often becomes a multi-year process of 'putting it into the medium-term management plan, getting approval in the next fiscal year's budget, etc.'

In Sierra's case, because they don't touch existing SoRs, short-term implementation becomes possible at the discretion of business-side department heads, such as the Head of Customer Support, Division Manager, or CX Executive. In fact, Sierra has a case study where 'a major health insurance company went live in 7 weeks from the start of the project' (Source: Sierra Year Two Blog). This is in stark contrast to the 'multi-year major projects' of ERP implementation in the SaaS era.

Point 2: Not attacking the stronghold of existing SaaS players head-on

If you go to 'replace' Salesforce, ServiceNow, or Zendesk, you take on the full set of existing contract terminations, data migration risks, and organizational resistance. Sierra does not replace them, but rather positions itself in the layer 'above' them. As a result, they avoid head-on collisions with existing vendors and, moreover, capture the 'final interface' for the customer.

Point 3: Directly capturing the evolution of model reasoning capabilities as the company's own competitive advantage

Designing to use reasoning instead of data integration means that the evolution of foundation model capabilities (GPT-4 → GPT-5 → Claude Opus → ...) manifests directly as an improvement in Sierra's product power. This is also consistent with the 'Constellation of Models (multi-LLM strategy)' described later.

2-3. Starting with the last analog channel: 'Telephone'

Another strategic 'contrarian' move by Sierra is their channel strategy.

Most chatbot players entered the market through web chat. While Sierra was also chat-centric in its early days, in October 2024, just eight months after launch, they released support for telephone (PSTN) interactions.

Why is the telephone so decisive? Bret's observation is sharp.

There is a huge implication here. By capturing the telephone, the long-divided 'digital team' and 'call center team' within client companies are being unified around Sierra.

And once an organization is unified, it is extremely difficult to return to the original state of division. This organizational integration itself becomes a switching cost for Sierra. With 'channel-specific add-ons' like Zendesk, you cannot trigger this kind of organizational transformation.

In fact, as of October 2025, Sierra has reached a state where 'Sierra handles more phone calls than chats' (Source: Sierra Blog "Voice Turns One"). The strategic decision to go after the telephone from day one became a driver in just one year.

2-4. Ensuring Reliability: Constellation of Models and Supervisor Agents

The biggest concern for enterprises when deploying AI into actual operations is hallucination. Sierra solves this with a multi-LLM architecture they call the 'Constellation of Models'.

Specifically, they employ a two-tier design that combines multiple LLMs (over 15 models, including OpenAI, Anthropic, Google, Meta, and their own open-weight models) where the output of a generative model is verified by another Supervisor Agent. As Clay Bavor says:

The data from τ-bench (Tau-bench, the de facto standard for agent evaluation) published by Sierra itself is shocking. In a return processing simulation, the success rate of a frontier LLM alone is only 61%, and 35% for flight reservation changes. The 8-consecutive-success rate (pass^k=8) drops to 25%. In other words, a single LLM is not usable at an enterprise level (Source: τ-bench paper, published June 2024).

The reason this architecture is strategically important is that it solves the barrier to entry for regulated industries (finance, healthcare) at the product level. Compared to products locked into a single foundation model like Salesforce Einstein, Sierra makes the flexibility of model selection itself a differentiator.

Part 3: The Ingenuity of the Monetization Model—Why Outcome-Based Billing is the Core of Sierra's Strategy

3-1. What is Outcome-Based Billing?

On December 10, 2024, Sierra became the first in the industry to officially announce Outcome-based Pricing (Source: Sierra Blog "Outcome-based Pricing"). The mechanism is simple.

• A fee is charged for every issue autonomously resolved by an AI agent

• If escalated to a human, it is free

• One theory suggests it is about $1.50 per resolution (Source: Sacra). This is about 10% of the cost of a human agent ($10–$20)

I believe this is not just a pricing model, but the core of Sierra's strategy.

3-2. Why the Customer Service Sector is a Perfect Match for Outcome-Based Pricing

Outcome-based pricing does not work in every sector. Among them, customer service is a field where the conditions for outcome-based pricing are almost perfectly met.

Areas where there are 'outcomes that can be measured independently, immediately, and binarily' will become the main battlefield for outcome-based AI agents. Sierra likely chose the call center sector precisely because there was a clear 'outcome-based pricing fit'.

(Other potential areas suitable for outcome-based pricing include: inside sales [number of meetings set], contract review [number of contracts], recruitment screening [number of candidates passing to interviews], and insurance assessment [number of assessments completed].)

3-3. Why This Becomes a 'Structural Moat' Against Existing Players

Bret Taylor speaks about this point very consciously.

What does this mean?

Salesforce, Zendesk, and ServiceNow all have shareholder expectations, sales organizations, and earnings forecasts based on seat-based pricing (number of users × monthly fee). If these existing giants were to declare, 'From now on, we will use outcome-based pricing,'

• existing seat license revenue would be significantly damaged in the short term,

• the evaluation metrics for sales organizations (number of contracts, MRR) would stop functioning,

• and they would face accountability to analysts, causing stock price volatility.

This is structurally difficult to do. That is precisely why it has become a unique position that startups can seize first.

In addition, the transition from seat-based pricing to outcome-based pricing triggers conversations from customers at every contract renewal that 'it has become cheaper than before.' For existing vendors, this is nothing but downward pressure on prices.

To borrow Bret's words:

Sierra's argument is that this is a bet with the potential to change the structure of the SaaS industry.

3-4. The Side Effect of Outcome-Based Pricing: 'Shifting Implementation Responsibility to the Vendor'

There is another important side effect to this outcome-based pricing.

In traditional SaaS implementation, the 'vendor (seller),' 'SIer (implementer),' and 'customer (user)' are fragmented, and to borrow Bret's words, "success has a thousand fathers, failure is an orphan"—a situation where everyone points fingers at others when things go wrong has become the norm.

With outcome-based pricing, this is structurally resolved. Since Sierra is 'charged based on the number of resolutions,' they have no choice but to take responsibility for everything themselves, from data cleaning and prompt adjustment to tuning supervisor models and designing on-site operations.

The result of this is Sierra's 'Forward Deployed Engineer' culture. This is a model established by Palantir in the 2010s and inherited by modern top AI companies like OpenAI, Harvey, and Ramp. Sierra calls this role 'Agent Engineer' and has even established a rotation program for new graduates (APX Program).

The conclusion this model leads to is:

• A new positioning called 'Productized BPO': A third entity that is neither software, consulting, nor BPO, which manages the customer's entire operations

• Switching costs that cannot be replicated by product alone: Customer business knowledge, journey design, and guardrail adjustments are all accumulated within Sierra

• Turning the difficulty of enterprise penetration into a 'weapon' rather than a 'barrier': The higher the barrier to entry, the harder it is for latecomers to imitate

※ Productized BPO = A form that replaces tasks previously handled by hundreds of operators with AI and software, adjusted by a small number of engineers in the background. It is a hybrid form that takes the core value proposition of 'business outsourcing' from BPO and the economics of 'scaling without adding headcount' from SaaS. It is a business model that is gaining traction in the US.

Part 4: From SoA to a New SoR—The True Identity of What Sierra is Building

4-1. What is the 'SoR of Processes' that Sierra is Accumulating?

If one were to describe traditional SoR in a word, it would be a 'record ledger.' Like CRM customer records or ERP journal entries, it is a passive database that statically holds 'what happened in the past'.

On the other hand, what Sierra is accumulating is fundamentally different in nature.

These are not 'records of what happened in the past,' but an accumulation of 'blueprints for how to run business operations going forward, their execution history, and a foundation for improvement'.

Bret Taylor himself speaks about this explicitly.

And,

Combining these two statements from Bret reveals his worldview.

The value of software changes in terms of 'how difficult it is to become obsolete' depending on what that software holds.The strongest are pure 'ledger'-type Systems of Record (SoR) like an ERP general ledger—accounting figures are difficult to replace, and their value persists for decades. Conversely, the most fragile are the UI and engagement layers (SoE)—their appearance is easily replaced by trends or new technologies. Between these two extremes, Bret inserts a new layer: '

Process SoR'. To use a Japanese analogy, it is like a 'comprehensive work manual' for business processes, as opposed to a traditional SoR (record ledger). However, it is a new, active, and dynamic type of SoR that holds not just manuals, but also execution history, customer relationships, and improvement logs all in one. And the truly important implication of Bret's point is this: the value of '

SoRs that are not pure ledgers' like CRM and customer support systems will shift to the agent layer that actually executes and records business processes—this is the logic by which Sierra can structurally siphon off value without having to replace existing SaaS head-on. Unlike traditional ticket history (past SoR) or CRM dashboards (engagement layer), an 'active SoR that defines and executes how to respond going forward' is being born here.

4-2. Agent Data Platform: Full-scale Productization of 'Process SoR'

On December 4, 2025, Sierra launched the Agent Data Platform (ADP) and announced that SiriusXM would be its first customer (Source: Sierra Blog "Agent Data Platform"). ADP is not

a repository of conversation logs, but a foundation that integrates relationships with each individual customer in chronological order. Specifically:

• Conversations across all channels (phone, chat, email, WhatsApp, on ChatGPT) consolidated per individual customer

• Accumulation of 'customer understanding' such as past churn attempts, complaints, preferences, and personality traits

• Data from existing systems such as CRM, billing, and ERP referenced in connection with the context of the conversation

• History linked to outcomes, such as the success or failure of resolution, subsequent satisfaction, and frequency of re-inquiries

In short, it is a foundation that continues to understand people across conversations. If a traditional CRM is a 'warehouse for sales information,' ADP is a 'command center for remembering relationships with those customers and acting optimally at the next touchpoint.' In line with this, at the

Sierra Summit 2025 in November 2025, Sierra formalized the expansion of its positioning from a 'customer service tool' to a 'happy customer machine' (an entity responsible for the entirety of customer acquisition, growth, and retention). It is expanding its business domain from one-off inquiry handling to the orchestration of the entire customer lifetime. The worldview Sierra is painting here is:

all customer touchpoints for a company will be conducted through Sierra. And the insights, relationships, and history generated at those touchpoints will all be accumulated within Sierra. While this is similar to how 'customer data gathered in CRMs' in the 20th century, the richness, activeness, and potential for utilization of the accumulated information are orders of magnitude greater.

4-3. How they are creating a structure where 'dependence on Sierra increases'

While building an SoR, it appears they are successively implementing mechanisms that make a world without Sierra untenable for customers. Let's look at the most recent representative initiatives.

① OpenAI Apps SDK Support (October 23, 2025): Aiming to capture the 'digital front entrance of companies'

On October 6, 2025, OpenAI announced the Apps SDK at DevDay. This is an SDK for building apps that run within ChatGPT, and launch partners include Expedia, Spotify, Zillow, Canva, Booking.com, and Figma.

Two weeks later, Sierra released a feature that allows agents created with Sierra to be published as apps on ChatGPT with a single click (Source: Sierra Blog "Publish to ChatGPT").

What does this mean? ChatGPT, with over 800M weekly active users, is transforming into a new digital gateway for brands to reach customers. You could say that what happened with Google being the gateway to the internet for 20 years is about to happen with ChatGPT.

At that point, companies will face the question: "Should we place our own agent on ChatGPT or not?" Companies that have already built agents on Sierra can deploy them to ChatGPT with a single click. Conversely, companies that have not yet adopted Sierra have the choice of either building a separate agent for ChatGPT or adopting Sierra to handle it with a single click.

This is a move to capture the position of "Sierra = the Single Source of Truth for a company's AI agents." Once this position is secured, Sierra can monopolize the "bridge" to any new channel that emerges.

② Level 1 PCI Compliance (April 7, 2026) — Toward an "Interface That Runs All the Way to Payment"

Customer service is the "gateway to support," but there have traditionally been security barriers to further operations (purchasing, payment, contracting). Credit card information and ACH payments require highly advanced PCI DSS Level 1 certification, and it was technically difficult for AI agents to comply with this.

In April 2026, Sierra achieved the first Level 1 PCI compliance as a conversational AI platform and released a feature that allows card/ACH payments to be completed within a conversation, both via chat and voice, without holds or transfers (Source: Sierra Blog).

This has huge strategic significance. Sierra's role is elevated from a "gateway to support" to a "commercial interface that runs all the way to payment". Cases like Rocket Mortgage, where Sierra's agents handle everything from "home search to loan origination to servicing," will spread across all industries in the future.

A new metric, "how much transaction volume is flowing through AI agents," may even become the center of a company's digital metrics.

③ Ghostwriter (March 25, 2026) — Automating Agent Construction Itself

They announced an agent-generating agent that "builds, executes, and continuously improves intelligent agents just by describing the outcome" (Source: Sierra Blog "Ghostwriter"). Ghostwriter automatically generates multi-language, multi-channel agents from SOPs, past call transcripts, whiteboard photos, voice recordings, and more.

This means "bringing the marginal cost of Sierra adoption as close to zero as possible," which triggers an "exponential expansion of usage scope" within customers. It supports the trend where companies that have adopted Sierra continue to deploy agents to new use cases (sales, onboarding, internal help desks, etc.) one after another.

④ Contact Center Integration and Live Assist — Swallowing Even BPO Operations

Agent OS 2.0, announced at Sierra Summit 2025 in November 2025, included the Live Assist feature. This is a feature that provides real-time guidance to human operators, where Sierra supports from behind the scenes when humans handle complex cases that AI cannot fully resolve.

This allows Sierra to cover both "cases solved by AI alone" and "complex cases handled by humans." It can be said that this is a move to bring areas traditionally handled by the BPO industry onto the Sierra platform.

4-4. Anatomy of the Lock-in Structure — Why Is This Dependency Structurally Robust?

The set of measures we have looked at so far creates a situation where it becomes "extremely difficult to operate modern CX without Sierra." If you calmly break down the lock-in mechanism, it has the following multi-layered structure.

① Organizational Layer Lock-in

By taking over the phones, the "digital team" and "call center team" are integrated within the customer company. The cost of re-segmenting an organization once it has been integrated is extremely high, and this in itself becomes a barrier to leaving Sierra (as discussed in Part 2-3).

② Operational Layer Lock-in

Through outcome-based billing, Sierra takes responsibility for implementation by being embedded on-site (Forward Deployed Engineer/Agent Engineer). The customer company's journey design, guardrail settings, Supervisor Agent tuning, and industry-specific know-how are all accumulated within Sierra. Transferring this to another vendor would incur reconstruction costs that could take months or years.

③ Data/SoR Layer Lock-in

The Agent Data Platform accumulates conversation history, relationship data, resolution patterns, and inference logs with customers in a format unique to Sierra. This differs in nature from traditional CRM data, and even if exported, the equivalent value cannot be replicated in other systems.

④ Channel and Distribution Layer Lock-in

Every time a new channel or capability is added—such as ChatGPT, contact centers, or payments—it can be handled centrally via Sierra. Conversely, switching to another vendor would require rebuilding all of these.

⑤ Asymmetry of Economic Rationality

From the customer's perspective, the option to build and operate all these functions in-house effectively does not exist:

• It is practically impossible to build phone, chat, WhatsApp, ChatGPT, and contact center integration in-house.

• Building an architecture that combines over 15 LLMs and monitors them with a Supervisor Agent requires an investment of hundreds of person-months.

• It is also unrealistic to independently obtain regulatory compliance for dozens of countries, PCI Level 1, and ISO/IEC 42001.

• There is no point in bearing the operational burden of continuously evaluating and incorporating foundation models that evolve every month.

In other words, an asymmetry is built in where the 'option to not depend on Sierra' is difficult to justify both technically and economically.

And this lock-in comes with clear results. WeightWatchers automated 70% of inquiries within one week of implementation (maintaining a CSAT of 4.5 or higher, Source: Sierra Customer Story), SoFi saw its NPS increase by 33 points after implementation (Source: Cheeky Pint), and Ramp is automating 90% of inquiries in 2025 (Source: ibid.).

Sierra's lock-in is designed as 'a dependency that cannot be structurally resolved in exchange for results.' Because the economic rationality for leaving does not exist, customers remain for the long term—is this not the essence of the moat that Sierra is building?

Part 5: Summary—What Sierra Has Shown to SoA Players in the AI Era

5-1. Strategic Pattern: The Three-Tier Structure of 'SoA → Interface → SoR'

Abstracting Sierra's success reveals an example of a strategic pattern for the AI era.

The essence of this pattern lies in inserting a new layer itself into the 'gaps' of the layer divisions protected by existing players.

By placing an active layer that operates through inference (SoA) between SoR (Salesforce/Zendesk, etc.) and SoE (various UI/UX), and using that as a base to capture channels and data, they ultimately build a new type of SoR. This is a position that is structurally difficult for existing players to follow, no matter how much capital and talent they have.

5-2. Sierra Success Requirements: The Four Prerequisites That Made This Strategy Possible

There were several structural prerequisites in the background that allowed Sierra to execute this three-tier structure.

Requirement 1: The domain must have outcomes that can be clearly defined by a single event and measured immediately (Part 3)—Customer service was a rare domain where 'resolved/not resolved' could be immediately determined as a binary outcome, and causality with other measures was easy to isolate. When launching a business in the application layer of the AI era, whether this question can be asked first is the starting point.

Requirement 2: A design philosophy that does not aim to disrupt existing Systems of Record (SoR) (Part 2)—Avoiding the quagmire of replacing existing SaaS and positioning oneself in a higher layer to be adopted at the discretion of department heads is the key to short-term PMF.

Requirement 3: Execution capability to capture the most organizationally deep channels first (Part 2-3)—By capturing high-difficulty channels like telephony early, organizational integration occurs on the customer side, which itself becomes a switching cost.

Requirement 4: Technical capability to turn model selection flexibility into a product (Part 2-2, Part 4-4)—The multi-LLM + supervisor model architecture is a permit to enter regulated industries and provides a meta-positioning where 'Sierra wins regardless of which model company wins.'

Only by meeting these four requirements did Sierra's three-tiered structure succeed. While there are differences depending on the domain, these should be important focal points for companies competing in the application layer in the AI era.

Closing: A new winning strategy for the AI era—building a new SoR starting from SoA.

Let's return to the opening question.

Why was Sierra able to build a position with a $10B valuation and over $150M in ARR in two and a half years in the 'AI for call centers' space—a red ocean already crowded with Zendesk, Salesforce, Genesys, Intercom, and others?

My answer is that Sierra was not playing the same game as the existing players.

Existing players were competing in the traditional vertical structure of 'SoR (Salesforce/Zendesk, etc.) → SoE/UI on top of that.' Sierra entered from the side. They established an SoA (System of Action, an agent layer that operates via inference), took a position that bypasses and references multiple existing SoRs, integrated organizations from the deepest channels, and as a result, built an entirely new 'SoR of processes'—it was a strategy of inserting a new layer itself into the gaps between existing layer divisions.

I would like to re-examine what is essentially interesting about this positioning.

First, it is structurally difficult for existing players to follow.

Just as Salesforce cannot say, 'From now on, we will charge based on outcomes,' it is difficult for existing giants burdened with shareholder expectations for seat-based SaaS to adopt the same business model and implementation responsibilities as Sierra. Sierra is not tearing down the strongholds of existing SaaS head-on; it is building a 'new castle' next to them.

Second, it proves that a domain that looked like a red ocean from a conventional perspective was actually a blue ocean.

The call center domain appeared to have many competitors, but no one was attacking it starting from SoA. In other words, now is the time when new domains that can only be captured due to the evolution of generative AI/LLMs are being born. Asking 'Is there a layer that existing players cannot defend?' rather than 'It's impossible because existing players are there' will be a new perspective for evaluating startups in the AI era.

Third, the sequence of entering from SoA to build an SoR itself becomes one of the winning strategies unique to the AI era.

Instead of going through a multi-year data integration project, bypassing existing SoRs with inference to go live in a few weeks, using that track record as a stepping stone to accumulate data, and building a new SoR—this time compression was impossible before LLMs.

And this strategy, to borrow Bret's words, is a position that will generate massive economic value in the AI era.

Even without waiting for the evolution of foundation models, there are already trillions of dollars in unrealized economic value lying dormant. The companies that will capture this are enterprise AI firms that embed themselves deeply into industries and business domains to sell "outcomes."One of the strongest candidates for this is Sierra.

Selling outcomes rather than productivity improvements, selling operations rather than software, and instead of replacing the System of Record (SoR), entering through the System of Action (SoA) to build a unique SoR—this is one of the classic patterns for breaking the stronghold of existing SaaS in the AI era.

The play pioneered by Sierra—"building a new SoR starting from SoA"—proves that even in areas that appear to be red oceans, there are new ways to win. I hope this article serves as a compass for deciphering those movements.

List of Reference Sources

• Bret Taylor, "Cheeky Pint with John Collison" (Stripe Press, published April 2026)

• Bret Taylor, "Latent Space Podcast" (February 2025)

• Clay Bavor, "Training Data" (Sequoia Capital Podcast)

• Sierra Blog: Year Two, Voice Turns One, Outcome-based Pricing, Publish to ChatGPT, Agent Data Platform, Ghostwriter

• CNBC, "Sierra valuation hits $10 billion" (2025/09/04)

• Sacra, "Sierra Company Profile"

• τ-bench paper (arxiv: 2406.12045)

• Sierra Customer Stories (WeightWatchers, Rocket Mortgage, SoFi, etc.)

(Note: The figures, dates, and facts mentioned in this article are based on information publicly available as of April 19, 2026. Furthermore, this article includes the author's personal views and analysis and does not represent the official position of any company or organization the author is affiliated with.)