AI Avatars Just Walked Into the Bank: Why Last Week's Citi Sky, Customers Bank AI Clone, and DeepBrain AI Studios Mark Financial Services' Avatar Moment

For the last six weeks the AI avatar story has crossed almost every line of business worth crossing. It walked into retail customer service with Cognizant and Google Cloud. It walked into the classroom at Boise State and UVA. It walked into Mayo Clinic exam rooms for patient education. It walked into Google Vids and Zoom at the workplace. It even walked into the $20 dropshipping ad at the SMB end of e-commerce.

Last week, AI avatars walked into the bank.

On April 22, 2026, Citi Wealth unveiled Citi Sky, an always-on AI-powered member of the Citi Wealth team built on Google Cloud and Google DeepMind technologies. Five days later, Customers Bank's CEO Sam Sidhu let his AI clone deliver the company's first-quarter earnings call — what he described as a first for a US public company — and announced a multiyear OpenAI partnership to automate lending and onboarding across the $25.9 billion asset bank. Two days after that, DeepBrain AI launched real-time AI Avatar Agents on its AI STUDIOS platform, backed by more than 100 on-device enterprise deployments — including Shinhan Bank and Samsung Securities.

Three different banks. Three different layers of the financial services stack. One week.

This is the moment the avatar layer arrives in financial services. It is also genuinely good news — for clients, for advisors, for the AI avatar economy underneath, and even for the institutions themselves. Here is why.

The Three Beats Of The Week

The headline is best understood as three discrete moves landing in the same seven-day window, each at a different layer of the bank.

The first beat is the client-facing avatar in wealth management. Citi Sky is the loudest version of this so far. Announced at Google Cloud Next 2026, it is a real-time AI avatar built on Google DeepMind's latest avatar technology and Gemini's live audio and video models, integrated into Citi Wealth platforms via the Gemini Enterprise Agent Platform. Sky launches in summer 2026 in a phased rollout to Citigold clients — clients who maintain an average monthly balance of $200,000 or more — beginning in English and Spanish. It will surface market insights from the Chief Investment Office, prompt clients about CD maturity events, and surface anticipated financial needs as the relationship deepens.

The second beat is the executive-presence avatar for institutional communication. Customers Bank, a Pennsylvania-based commercial lender that has grown significantly under Sam Sidhu, used his AI clone to deliver the prepared-remarks portion of the Q1 2026 earnings call — apparently the first time a public US company has done so. Paired with the announcement is a multiyear OpenAI deal in which OpenAI engineers embed inside the bank to automate lending, document collection, and client onboarding. The bank is targeting commercial loan close times falling from 30–45 days to about seven, and complex business account opening collapsing from a full day to under twenty minutes.

The third beat is the on-device avatar at the branch and kiosk. DeepBrain AI's AI STUDIOS platform released real-time avatar agents engineered to run on-device — banking kiosks, digital signage, tablets, and frontline mobile devices — with native multilingual support across more than 150 languages and a model-agnostic architecture that connects to commercial AI services, open-source models, or a bank's own in-house LLM. The footprint already includes more than 100 enterprise deployments, with Shinhan Bank and Samsung Securities named explicitly in the financial services slice.

Each of these moves would be notable on its own. Together, in one week, they describe a fully populated avatar layer for a regulated industry that has, until now, mostly watched the AI avatar wave from the shore.

What Citi Sky Actually Is

Strip the press-release language and Citi Sky is doing three things that traditional digital banking apps could not do.

The first is conversational fluency at the wealth-tier complexity ceiling. A Citigold client does not have a question that fits inside a chatbot's intent tree. They have questions like "what does the new tax framework mean for my muni allocation," "should I roll my expiring CD into duration or stay floating," and "what is our house view on emerging-markets equity right now." Sky is built specifically to surface market views from Citi Wealth's Chief Investment Office in real time, using Gemini's live audio and video models. That is the first time the actual asset of a wealth franchise — its CIO's research output — has been deliverable through a 24/7 avatar interface that talks back.

The second is anticipation, not reaction. Sky is described as designed to anticipate financial needs and surface prompts proactively — for example, alerting a client to a CD maturity event before the client thinks about it. The avatar interface is the surface; the underlying win is that the bank's data systems can now reach the client through a face and a voice that the client will actually engage with, instead of through an email the client will not open.

The third is a phased, regulated rollout. Sky is launching in English and Spanish first, scaling languages over time, available to a clearly defined Citigold tier, in a phased US rollout. That is what responsible deployment looks like in a regulated industry — and it is the template every bank is going to copy.

The win for clients is that the experience of being a wealthy client of a global bank — real-time access to market insight, proactive guidance, fluent two-way conversation — finally extends to the moments between advisor calls. The win for advisors is that Sky absorbs the routine inquiry load, freeing up advisor time for the conversations that actually compound trust and AUM.

What The Customers Bank Move Means

The Customers Bank story is the one most people will remember from last week, because letting an AI clone of the CEO read the prepared remarks of a public company's earnings call is the kind of move that gets covered by Fortune and CNBC on the same day.

But the more interesting story is the operational one underneath. Customers Bank has signed a multiyear OpenAI partnership in which OpenAI engineers embed inside the bank to deploy AI across commercial banking. The targets are concrete and measurable: collapse a 30–45 day commercial loan close to about seven days; collapse a one-day business account opening to under twenty minutes; move the bank's efficiency ratio from roughly 49% to the low 40s, with returns improving from 2027 onward.

Two things are worth pulling out.

The first is that the AI clone on the earnings call is not a stunt — it is a thesis. Sidhu's argument, made through the medium itself, is that an authoritatively trained avatar can stand in for a senior leader on a high-stakes communication, with full disclosure, and the world keeps turning. If it can do that, then most of the bank's middle layer — the loan officer reading the same product script, the compliance officer reading the same disclosure, the onboarding specialist explaining the same KYC checklist — is also liberatable. That is what the OpenAI partnership is built to operationalize.

The second is that 75% of Customers Bank employees already use ChatGPT Enterprise, the deployment having begun in 2023. The earnings-call avatar is not the start of the journey; it is the public marker of a multi-year internal saturation. The lesson for other US banks is that the avatar moment is a leading indicator, not a starting line. Institutions that have been quietly building AI fluency since 2023 have a runway in 2026 that institutions still in pilot do not.

What DeepBrain AI's On-Device Avatars Unlock

DeepBrain AI's launch is the quietest of the three but, structurally, may be the most important for the long arc of the financial services rollout.

The platform's distinguishing feature is on-device inference. The avatar — and the LLM behind it — runs locally on the kiosk, the tablet, or the mobile device. There is no round-trip to a cloud endpoint for every conversational turn. That single architectural choice changes three things at once for a regulated industry.

It collapses latency to a level where natural conversation is genuinely possible — no awkward two-second pause between the customer asking and the avatar answering, which is the gap that has historically broken on-screen kiosks at branches and airports.

It simplifies the data residency story. The customer's question, the avatar's response, and the underlying inference all happen inside the device on the bank's premises. For institutions navigating EU, Korea, Japan, Brazil, and US-state-level data localization rules, that is a decisive deployment advantage.

And it is model-agnostic. The bank can choose the LLM stack — commercial, open-source, or proprietary — that fits its risk and compliance posture. Shinhan Bank does not have to use the same model as Samsung Securities, even if both deploy AI STUDIOS. That flexibility is non-negotiable for the second wave of financial-services adopters who are not going to bet on a single foundational model vendor.

Across the existing footprint, AI STUDIOS is already serving banking kiosks in lobbies, retail signage on store floors, healthcare tablets in hospitals, and frontline mobile assistants. Adding more than 100 deployments before formally launching the real-time agent product gives DeepBrain AI a distributional advantage that competitors will spend the rest of 2026 trying to close.

Why This Is Good News For Clients

It is reasonable to greet "AI avatar enters your bank" with a healthy dose of skepticism. We have all sat through a chatbot that could not understand a basic refund question. So it is worth being specific about what changes for the client.

Wealth-tier clients get research access on demand. Citi Sky surfaces the Chief Investment Office's market view in real time, in conversation, in the client's preferred language. That is a tier of access that previously required a scheduled advisor call or a quarterly newsletter. The client now gets it at 9pm on a Sunday, in the same medium they use to talk to friends.

Mass-market clients get fluent on-device service. A DeepBrain AI–powered avatar at a branch kiosk can answer a complex account question, in 150+ languages, without referring the client to a queue. For first-generation immigrants, multilingual professionals, and travelers, that is a meaningful uplift in the basic dignity of the bank-branch experience.

Commercial clients get speed. Customers Bank is targeting a 7-day close on commercial loans that previously took 30–45. For a small business waiting on working capital, that is the difference between making payroll and missing it, between buying inventory in time for the season and missing it. The avatar layer plus AI automation underneath turns the bank into something that responds at the speed of the business it is funding.

Every client gets proactive engagement. The CD that quietly matured into a checking account at zero yield. The 401(k) rollover that never happened. The mortgage refinance opportunity that came and went. The avatar layer turns these from missed opportunities into prompted conversations, with full client agency over whether to act.

These are real consumer wins. They are not theoretical.

Why This Is Good News For Advisors

A common pushback whenever AI walks into a high-skill profession is that the human role is being squeezed. The shape of last week's announcements actually points the other way.

Citi has been explicit that Sky is built to work alongside financial advisors, not replace them. The avatar handles the constant, low-margin inbound — the price check, the maturity reminder, the balance question, the scheduled prompt — that absorbs disproportionate advisor time without producing relationship value. The advisor, freed of that load, can spend more time on the high-trust conversations that actually compound the book — estate planning, generational wealth strategy, philanthropic structuring, the moments where a client's life event changes what their portfolio should look like.

The same logic applies inside Customers Bank. A loan officer whose 30-day file just collapsed to seven now has time to bring in three new commercial relationships during the same period. A relationship manager whose onboarding workflow just dropped from a day to twenty minutes can run more onboardings, attend more client visits, and bring more institutional knowledge to bear on actual structuring decisions.

The pattern across the three announcements is consistent: the avatar absorbs scheduled, repetitive, and reactive work; the human absorbs unscheduled, complex, and proactive work. That is the right division of labor, and it is the one good firms have been trying to engineer for two decades. AI avatars finally make it operationally possible.

Why This Is Good News For The Avatar Layer Underneath

Zoom out and last week's announcements are arguably as significant for the AI avatar economy as any enterprise news in 2026.

Financial services is the last industry to adopt new computing categories, because the regulatory cost of getting it wrong is enormous. The fact that three discrete financial institutions — a global wealth franchise, a US commercial bank, and two Korean financial giants — all stood up real-time avatar deployments inside one week tells the rest of the market that the compliance, risk, and data-residency questions have been answered well enough for tier-one institutions to commit publicly.

Once banking adopts a category, every adjacent regulated industry follows. Insurance, asset management, real estate brokerage, wealth advisory networks, regional credit unions, neobanks — the procurement-cycle clock starts now, not in 2027.

That means demand for high-quality, identity-faithful avatar inference is about to compound. Citi Sky, the Customers Bank executive clone, and the DeepBrain AI on-device fleet all sit downstream of a small set of foundational avatar engines — Google DeepMind's real-time avatar tech, OpenAI's Sora-derived video stack, Synthesia and HeyGen on the enterprise side, and a growing tier of open-source avatar models that broke ground earlier in April. Every Citi Sky session, every Customers Bank operational avatar moment, and every kiosk conversation is incremental volume on those underlying models. Better unit economics drive the next price compression, which drives the next wave of adopters.

It also clarifies where the differentiation is moving. Once the underlying inference is commodity, brand-aligned avatar identity becomes the moat. Citi Sky needs to feel like Citi. The Customers Bank executive clone needs to feel like Sam Sidhu. The Shinhan Bank kiosk avatar needs to feel like Shinhan, not like a generic library template. The institutions that win the avatar layer are the ones whose avatars feel native to the brand — which is the work that platforms like Befamous.AI are built for, turning a company's actual identity, voice, and visual style into avatars that feel like the brand, not like a stock template the firm down the street is also running.

Three Other Beats That Reinforce The Pattern

Three quieter signals from the same week deserve a mention.

Google announced its most realistic avatars yet at Cloud Next '26, with the Gemini Enterprise Agent Platform positioning Gemini as the substrate underneath Citi Sky and a long list of customer deployments. The avatars now support custom branding — company logo on a shirt, fully branded backdrop — and will scale to 24 languages. That is the building block layer for the next 100 Citi-Sky-style deployments.

Customers Bank disclosed that 75% of its employees already use ChatGPT Enterprise, as the foundation under the new OpenAI deal. The announcement reframes how observers should read other banks' "AI strategies" — the live question is no longer whether the bank has tools, it is what percentage of the employee base has integrated them into daily workflow.

DeepBrain AI's footprint spans not just banks but retail floors, hospital tablets, public-service centers, and educational institutions like Kyung Hee Cyber University. The cross-vertical reach matters because it means the cost of building, training, and improving the on-device avatar stack is amortized across multiple industries. Banks benefit from work originally done for retail; retailers benefit from latency improvements driven by branch deployments. The long-run unit economics of the avatar layer underneath financial services keep getting better as a function of the deployments outside it.

What A Smart Financial Institution Will Do In The Next 90 Days

The playbook for a bank, wealth manager, insurer, or credit union watching this week is unusually concrete.

Start with the lowest-risk, highest-ROI surface: customer service and FAQ deflection. An avatar fluent in the institution's policy documents and product catalog, deployed first to the website and the mobile app, then to branch kiosks. Multilingual from day one. Use a model-agnostic stack so the LLM choice can evolve without rebuilding the avatar.

Move to proactive engagement for the existing book. CD maturities, loan refinance opportunities, paycheck-driven savings prompts, IRA contribution reminders. The avatar layer is the natural surface for proactive conversations the email channel has stopped delivering.

Layer in wealth-tier or commercial-tier real-time advisory, the Citi Sky equivalent for the institution's top client cohort. Build the avatar's tone, identity, and disclosure language in collaboration with the advisor, compliance, and brand teams.

In parallel, ramp internal executive and operational AI fluency — Customers Bank's lesson. Deploy ChatGPT Enterprise, Claude Enterprise, or Gemini Enterprise to the employee base. Track adoption percentage. The institution that hits 75% employee usage of an AI assistant in 2026 will, by 2028, look like a different company than the one that did not.

The institutions that take these four steps in the next ninety days will not be in the same competitive position in 2027 as the institutions that wait. Last week was the data point that confirmed it.

The Bigger Picture

Trace the AI avatar arc this year and the through line is hard to miss. Avatars showed up in advertising and creator content. They walked into the classroom and the exam room. They moved into the workplace and the retail front desk. They democratized down to the $20 dropshipper UGC ad. And last week they walked into the bank — at the wealth tier with Citi Sky, into the executive comms layer with Customers Bank, and onto the branch floor with DeepBrain AI's on-device fleet.

That is the line that matters. The same models, the same avatar libraries, the same conversational engines that power a $20 SMB ad on TikTok are powering a Citi Wealth conversation about a multi-million-dollar municipal bond ladder. The technology does not care which end of the market it is being used by. The market structure is what determines who benefits and when — and last week the structure moved decisively in favor of bank clients, advisors, and the AI avatar economy underneath.

Two years from now, the wealth client who started using Citi Sky in summer 2026 will have a different experience of being a Citi customer than the one they had in 2025. The Customers Bank borrower will close in a week instead of a month. The frontline branch teller in Seoul, São Paulo, or San Diego will speak the customer's language in real time, fluently, on-device. The AI avatar economy underneath will be larger, faster, and more identity-aware than it was the week before.

That is good news for clients. It is good news for advisors. It is good news for the avatar economy. And it is good news for the institutions that take it seriously this quarter — instead of two quarters from now, when the gap to the leaders is no longer closeable.

The avatar walked into the bank last week. The next twelve months belong to the banks that hand it a desk, a name, and a brand that feels like theirs.

Frequently Asked Questions

What is Citi Sky and when does it launch?

Citi Sky is an always-on AI-powered member of the Citi Wealth team built on Google Cloud and Google DeepMind technologies. Citi Wealth unveiled Sky on April 22, 2026 at Google Cloud Next '26. It uses Google DeepMind's real-time avatar technology and Gemini's live audio and video models, integrated through the Gemini Enterprise Agent Platform. Sky launches in summer 2026 in a phased US rollout to Citigold clients (those with average monthly balances of at least $200,000), beginning in English and Spanish, with additional languages to follow.

What did Customers Bank announce on April 27, 2026?

On April 27, 2026, Customers Bank — a $25.9 billion asset commercial bank — announced two things at once. CEO Sam Sidhu revealed that the prepared remarks on the bank's Q1 2026 earnings call were delivered by his AI clone, what he described as a first for a US public company. Simultaneously, the bank announced a multiyear partnership with OpenAI in which OpenAI engineers embed at the bank to automate lending and client onboarding. Targets disclosed include collapsing commercial loan close times from 30–45 days to about seven, and complex business account opening from over a day to under twenty minutes.

What did DeepBrain AI launch on April 29, 2026?

On April 29, 2026, DeepBrain AI launched real-time AI Avatar Agents on its AI STUDIOS platform. The product is engineered for on-device deployment — banking kiosks, digital signage, tablets, and frontline mobile devices — with native multilingual support across more than 150 languages and a model-agnostic architecture that supports commercial, open-source, or proprietary LLMs. The launch was backed by more than 100 enterprise deployments already in production, including Shinhan Bank and Samsung Securities in financial services, plus retail, healthcare, and public-sector customers.

Why is on-device avatar inference important for banks?

On-device inference matters for three reasons. First, it collapses latency to a level where natural conversation is possible — no two-second cloud round-trip per turn. Second, it simplifies data residency, because the customer's question, the avatar's response, and the underlying inference all happen on the bank's hardware. That is decisive in EU, Korean, Japanese, Brazilian, and US-state-level localization regimes. Third, on-device deployment lets the bank choose its own LLM stack, which is non-negotiable for institutions that do not want to be locked into a single foundational model vendor.

Will AI avatars replace financial advisors?

No, and the announcements from last week make the opposite case. Citi has been explicit that Sky is built to work alongside advisors, not replace them. The pattern is the same across the announcements: the avatar absorbs scheduled, repetitive, and reactive work — balance questions, CD maturity prompts, policy explanations, multilingual FAQs — while the advisor focuses on the unscheduled, complex, and proactive work where human trust compounds. The result is more advisor time on high-value relationships, not less advisor employment.

What is the Gemini Enterprise Agent Platform?

The Gemini Enterprise Agent Platform was announced at Google Cloud Next '26 as Google's one-stop platform to build, scale, govern, and optimize autonomous AI agents. It includes a no-code Agent Designer, infrastructure for real-time avatars, and integration with Gemini's live audio and video models. Citi Sky is one of the first publicly disclosed enterprise deployments built on the platform, and Google has confirmed that nearly 75% of Google Cloud customers now use its AI products to power production workloads.

How quickly can other banks deploy AI avatars?

Faster than most expect. Three deployment surfaces are immediately accessible. Customer service and FAQ deflection on web and mobile is the lowest-risk starting point. Proactive engagement for the existing book — CD maturities, loan refinance prompts, savings nudges — is the next layer. Wealth-tier or commercial-tier real-time advisory, the Citi Sky equivalent, is the highest-value layer. A bank that begins customer-service deployment in Q2 2026 can have a phased proactive-engagement layer in Q3 and a top-tier advisory deployment piloted by Q4. The institutions that compress this timeline will be in measurably different competitive shape entering 2027.

What is brand-aligned avatar identity and why does it matter?

Brand-aligned avatar identity is the practice of training an AI avatar so that its appearance, voice, vocabulary, and demeanor feel native to a specific institution rather than to a generic template library. As underlying avatar inference becomes commodity, identity becomes the moat. Citi Sky needs to feel like Citi. A Shinhan Bank kiosk avatar needs to feel like Shinhan. A wealth advisory firm's avatar needs to feel like that firm's particular tone of voice with clients. Platforms that turn brand identity into avatar identity — including Befamous.AI — are positioned to be the layer between the foundational avatar models and the bank's customer-facing surface.

Are AI avatars in banking safe and compliant?

Last week's announcements suggest the answer is yes, when deployment is staged and disclosed properly. Citi Sky is rolling out in a phased manner to a clearly defined client tier in two languages first, with explicit collaboration with the wealth advisory function. Customers Bank disclosed the AI clone's role on the earnings call openly, and the OpenAI partnership is structured around embedded engineers working inside the bank's compliance perimeter. DeepBrain AI's on-device architecture inherently keeps customer data on the bank's hardware. None of these are post-hoc safety stories — they are deployment patterns engineered for a regulated industry from the start.

What is the takeaway for a small wealth firm or community bank?

Three things. First, the cost of standing up an avatar layer has come down materially in 2026 — model-agnostic platforms like AI STUDIOS, plus brand-identity layers like Befamous.AI, are now accessible to firms that cannot build their own from scratch. Second, the right starting point is operational — a multilingual customer-service avatar on the website and mobile app — not a consumer-facing flagship like Citi Sky. Third, the institutions that move in the next 90 days will be in measurably better competitive shape in 2027 than peers who wait. Last week was the data point that confirmed it; the next twelve months will reward the operators who treat the announcement as a call to action rather than a news item.

This post is part of an ongoing series on how AI avatars and AI UGC are reshaping advertising, commerce, work, learning, healthcare, retail, and now financial services in 2026. Want to give your bank, wealth manager, credit union, or financial services firm a brand-aligned avatar that feels native to your institution rather than a stock template? Try Befamous.AI.

Sources