Unique Research · Insight / AIGC Open Day · September 8, 2023 · Theme: AI + finance · Guests: Youlianyun, Magic Engine, Feidi Tech — full roundtable rewrite
Open Day series (twice monthly) hosts founders/execs on frontier AI×industry. This session: finance’s shift from tech curiosity to business value · LLM multi-domain outlook · commercialization challenges · 1–3 year outlook.
Youlianyun (有连云) — “everything connected”
Yunchuang — AI content creation for real-time market follow (stocks/funds move too fast for pure manual). Yunlian — distribute generated content onto finance platforms for sales/marketing assist. Yunshu — content-effect monitoring + analytics for decisions. Accuracy stakes are extreme; generic LLMs often insufficient → proprietary finance models. Clients: fund-industry leaders plus listed semis/new energy/biopharma; products already solve sales/marketing pain and guide roadmap.
Magic Engine (魔数智擎)
Roots in first-gen data-mining vendor Salford Systems; China core team founded 2018. One-stop financial intelligent decision + AIGC LLM platform: lower model barriers from data→modeling→production for risk control, precision marketing, smart ops. “Last mile” of model landing must be fluent and business-fit—quality, explainability, value match. LLM-era products mimic human data/corpus handling; speed coding for analysts; offline/private deploy for bank risk. Clients across banking/insurance/securities/healthcare/auto/energy (~7 industries; ~50 financial institutions). Cases: first domestic bank project putting complex ML models into credit-card A-card real-time decisioning; state bank VIP churn −76.04% and +¥250M wealth-product sales; regional bank—2 people / 3 months / 10 models → +¥1.1B AUM.
Feidi Tech (飞笛科技)
Digital AIGC for financial institutions’ user-growth data engine; partners brokers/funds/banks. Full-lifecycle trust management: acquire → retain → activate → convert. Team: senior financial media + internet PMs + quant experts. Services: online “virtual advisors” for product teams; AIGC marketing-material platform for ops; “businessized” digital ops middle platform for data/AI units; professional wealth plans + online content companionship for advisors to win investor trust.
Roundtable
Attitude: tech → business value
Zhang Yan (Youlianyun): Finance is relatively market-open to new tech—but tech is a tool; must solve real needs, not exist for its own sake; accelerate business as AI marketizes. Xu Jing (Magic Engine): Two sink phases—post-2015 head institutions saw AI value; lately recognition sinks to mid/small FIs via marketing + head cases. New bar: no more trial-and-error tolerance; solutions must create lasting business value chains with clients. Qiu Huihui (Feidi): vs prior wave, this wave stresses business pressure (stock-client activity down); FIs urgently need problem-solving; product+tech+biz collaboration for sustainability—not tech theater.
Where LLMs unlock new products
Zhang: LLM as methodology across finance/weather/power/pharma; fast landing in CS; extend to risk, AML. Qiu: stock efficiency (integrate LLMs for data quality/speed) + incremental apps (buyer-side financial info; finer scene granularity + vertical data; better interaction / noise reduction)—needs ecosystem partners; open-boundary disagreements still block data/tech collaboration. Xu: domain data → high-accuracy solutions (Huawei Pangu weather/manufacturing; power load; bank-statement vertical models for AML/fraud); multimodal + traditional data next.
Commercialization challenges
Qiu: ToB clients push short-term monetization vs long vision—align client ROI with company vision; user-centric design; predictable outcomes; drop costly misaligned projects. Xu: choose staffing/product/consulting/ecosystem positioning; product-core → ecosystem wins ToB+AIGC. Zhang: do long-value work; bind product to needs; iterate; profitability-first for sustainability.
1–3 year outlook
Zhang: keep landing AI with multiparty ecosystems; focus rigid needs/pain; avoid high-risk adjacent bets. Qiu: more important than tech novelty—business-model innovation in a licensed industry: transaction-centric → user-centric digital transformation; platform/ecosystem experiments for better service + resource integration. Xu: finance is conservative; validation cycles long; near-term innovation stays in mature lanes (analytics, risk, marketing, fraud); disruptive leaps slower.
CN images; attach on ingest.
— END —