Apple Podcasts  ·  Spotify  ·  Buzzsprout

Paper A

The Influence of AI-Driven Marketing on the Financial Performance of Digital Banks in Nigeria

Asanga Aniekan, Okon Akpan Aniefiok — 2026 — Zenodo (CERN European Organization for Nuclear Research)

peer reviewed journal article  ·   ·  read now

Key findings

• AI tools that tailor banking experiences to individual customers — like showing you relevant offers or account tips — explained about 61% of the variation in bank financial performance (R² = 0.611).
• AI-powered CRM systems — tools that help banks track, manage, and improve relationships with customers — were the strongest predictor of financial performance, explaining about 66% of the variation (R² = 0.663).
• AI chatbots and virtual assistants (the kind you chat with when you have a banking question at 2 a.m.) also positively predicted financial performance, explaining about 57% of the variation (R² = 0.572).
• All three AI marketing tools showed a statistically significant positive relationship with financial performance — meaning banks that used them more appeared to perform better financially.

Marketing implications

• If you work in fintech or digital banking marketing, this study suggests AI-powered CRM — tools that track customer behavior and help you reach out at the right moment — may be the single highest-leverage AI investment for revenue. Prioritize getting that right before adding chatbots or personalization layers.
• If your bank or fintech already has a chatbot, check whether it's actually solving customer problems or just deflecting them. The study links chatbot quality to financial outcomes — meaning a bad chatbot could hurt as much as a good one helps.
• If you're building a business case to invest in AI personalization (like recommending products to specific customers), this paper gives you a concrete stat to cite: a similar study found personalization alone explained 61% of the variance in bank financial performance in a comparable emerging-market context.

Paper B

The Dynamic Verifiable Multi-Agent Human Agentic Loyalty Loop (DVM-HALL) Model and the Net Human-Agent Score (NHAS) in Autonomous Commerce

Sai Srikanth Madugula, Peplluis Esteva de la Rosa, Daya Shankar — 2026 — arXiv

preprint  ·   ·  test this week

Key findings

• Traditional loyalty programs — built around human emotions, satisfaction, and brand attachment — break down when an AI agent is doing the buying. The AI doesn't feel brand love; it picks the option that best matches a set of rules and past performance data.
• The paper proposes that brand choice by an AI agent is shaped by five things working together: (1) how much the human user emotionally values a brand, (2) how well the AI has experienced dealing with that brand in the past, (3) how much the human trusts the AI, (4) how much decision-making power the human has given the AI, and (5) how reliably and verifiably transactions with that brand execute.
• The authors introduce a new score called NHAS (Net Human-Agent Score) — think of it like a 'trust rating' that tracks whether the AI agent's purchasing behavior actually matches what the human user wanted, based on real transaction logs and feedback.
• For brands operating in crypto or blockchain-based loyalty programs, factors like transaction fees, price slippage, and smart-contract security risks directly affect whether an AI agent will pick that brand — these technical costs become marketing factors.

Marketing implications

• If you run a loyalty program, start thinking about whether your rewards and brand signals are legible to an AI agent — not just a human. An AI shopping assistant won't feel nostalgic about your brand; it will optimize for price, reliability, and past performance. Make sure your brand data (ratings, transaction speed, return policies) is clean and machine-readable.
• If you're building a loyalty program that uses points, tokens, or any crypto-adjacent mechanics, the technical execution costs (transaction fees, reliability) will become as important as the emotional appeal of the rewards — because AI agents will factor those in.
• Start auditing what data your AI tools (recommendation engines, shopping assistants) are actually optimizing for. If a customer delegates purchasing to an AI, your marketing needs to influence the AI's criteria, not just the human's feelings.

Paper C

Hindcast: Replaying Prediction Markets to Evaluate LLM Forecasters

Xiao Ye, Jacob Dineen, Evan Zhu, Shijie Lu et al. — 2026 — arXiv

preprint  ·   ·  watchlist

Key findings

• Current AI forecasting tests are broken: when you ask an AI today to predict something that already happened (like who won the 2022 World Cup), it just recalls the answer from its training data — it's not actually predicting anything. This paper shows that problem is real and widespread.
• The HINDCAST system fixes this by pretending the AI only knew what was publicly available on Reddit before a specific past date — effectively putting the AI 'back in time' to test real forecasting ability.
• When given access only to pre-event Reddit posts (and no future information), AI retrieval tools still helped 8 out of 9 models make better predictions — so retrieval genuinely helps forecasting even without cheating.
• However, if the Reddit archive before the cutoff date only contained guesses and speculation (not real pre-event discussion), giving the AI that information made its predictions worse, not better. Noisy, speculative sources hurt AI judgment.

Marketing implications

• If you're using an AI tool to predict market trends, campaign outcomes, or consumer behavior, don't trust its 'forecasts' if it was trained or has access to data that already includes those outcomes. You may just be getting the AI's memory, not genuine prediction.
• When evaluating any AI forecasting product for business use (e.g., demand forecasting, trend prediction), ask vendors: 'How do you prevent the model from looking up the answer?' If they can't answer that clearly, treat its predictions skeptically.
• If you're building a marketing analytics tool that uses AI to forecast campaign results or industry trends, this paper gives you a methodology (frozen time-stamped data sources + prediction market benchmarks) to actually test whether your AI is forecasting vs. recalling.

Apple Podcasts  ·  Spotify  ·  Buzzsprout