Why is an ad agency building AI? — The natural next step for integrated marketing

Customer Service 2026-05-29 · Satsuma Creative · 10 min read

Satsuma is an ad agency, so why build AI customer service? Because the last mile of the ad funnel — the conversation after a visitor lands — has been outsourced to SaaS tools disconnected from the brand. We're taking it back, and growing AI teammates from the same logic as TVCs, social, and media buying.

TL;DR

  • We're an ad agency, not a SaaS — AI teammates are an extension of the ad funnel, not another SaaS competitor
  • The last mile of the ad funnel:the conversation after a visitor landshas been outsourced to SaaS tools disconnected from the brand. We're taking it back in-house.
  • Three concrete pain points: tone doesn't match the ad, product talking points differ from the TVC, hot leads come in with no one to catch them
  • Why other ad agencies don't do this: the technical leap is too big, the business model conflicts, the org structure isn't right
  • The core difference from SaaS:AI customer service is just the surface — what we're really selling is "a complete brand conversation strategy"
  • What it means for you: if you already work with an agency on advertising / social, having them handle AI customer service too (same tone) is more cost-effective than signing up for another SaaS

Short answer:Because the last mile of the ad funnel has been outsourced to SaaS tools disconnected from the brand.。

The long answer is below.

We're an ad agency, not a SaaS

Let's nail this down up front.

Satsuma Creative was founded in 2020. Our core business isintegrated marketing— TVCs, social, media buying, web applications. Our founder spent 20 years in the game marketing industry. Wedon't come from an engineering background, and wedon't sell software。

But since 2024, we've been building out the AI teammate product line.Why?

Not because "AI is hot" (that's bandwagon-jumping — Satsuma doesn't do that).We started because we noticed a missing piece in the ad funnel— and that missing piece is exactly what SaaS vendors do poorly.

The last mile of the ad funnel

Clients who work with us on TVCs + social + media buying mostly follow this flow:

1. 策略 + 創意(薩摩做)
2. 拍片 / 製作素材(薩摩做)
3. 媒體投放 / 社群操作(薩摩做)
4. 客戶看到廣告,點進網站
5. 跟網站對話 / 提問 / 留訊息(誰做?)←  ← ← ← 缺這塊
6. 業務跟進、close

Step 5Has long been outsourced to generic AI customer service SaaS— each client's needs handled in isolation, brand tone misaligned, wrong answers leak the funnel.

We spent 18 months optimizing steps 1-3 for clients, only to lose it all at step 5.

→ This isn't because SaaS vendors are bad— the SaaS model was never meant to manage "brand tone." Their business is selling standardized tools,not extending your brand universe。

Three concrete pain points

Not abstract theory. What we've seen on the ground with clients:

Pain point 1: The character from the TVC disappears the moment you hit the website

We shot a TVC for a game brand, the lead character was"a bard in a run-down tavern"— speaking in an archaic, wuxia-flavored tone.

The ad airs, players are drawn in,they click through to the official site, and a chat box pops up in the bottom right:

"Hello, I'm XX Company's smart customer service assistant. How may I help you?"

Completely breaks the ad's world. Players are yanked from "wuxia story" back to "call center" in an instant.

→ The atmosphere built by a NT$30 million ad budgetgets destroyed by a SaaS chatbot in three seconds.

We asked the vendor: "Can the chatbot also say something like 'Greetings, traveler, I am...'?" Vendor: "Sorry, the tone is set by the system default. We can't customize it for a single client.。」

→ That's the structural limitation of the SaaS model.。

Pain point 2: The chatbot can't answer questions about the campaign the ad promised

During the Lunar New Year media push, the SaaS chatbot's preloaded FAQ was from the previous campaign.

Player: "Does the Lunar New Year event end on 5/14?" SaaS chatbot: "For event details, please refer to the official site." (basically no answer)

→ The polished IG / FB ads pull people in,and the last mile feels like hitting a wall。

Fix: every new campaign requires the client tolog into the SaaS backend themselves and update the FAQ. But no one on the client side remembers, the marketing team only cares about the ad, customer service never gets the memo → every campaign, the same thing happens。

→ The problem isn't that "the client is lazy" — it's that SaaS pushing KB maintenance onto the client is structurally wrong in the first place. Clients pay for SaaS, not to hire another person to maintain the FAQ.

Pain point 3: Ad data and customer service data are siloed, no way to attribute the funnel

Ad path: FB ad → landing page → click chat → ask a question → leave email

As the ad agency, weneed to know: - Which ad creative attracted the people with the highest question rate? - Which keyword brought in people most likely to leave their email? - Where in the funnel do we lose the most people?

But the SaaS chatbot won't share that data with us(because the SaaS serves the client directly, not the ad agency).

We're stuck with Google Analytics + manually-exported Excel from the client + guesswork.The funnel is always half-blind。

→ ad strategy can't iterate precisely.

We chose to take this back in-house

After seeing these three pain points, there are two paths:

Path A: negotiate with existing SaaS vendors for "API access, custom tone, data sharing" → we tried, no one was willing, becauseit would break their SaaS business model

Path B: build it ourselves — make thispart of Satsuma's product line, alongside TVCs, social, and media buying

We chose B.

Why:Ad agencies already do "character design"(TVC leads),"brand tone"(social voice), and"funnel analysis"(media performance). The three things AI customer service needs — weknow them better than engineering firms do。

→ It's not "an ad agency forcing itself into AI." It's that "this work belonged to advertising all along" — it just got intercepted by the engineering crowd along the way.

Why don't other ad agencies do this?

Fair question. The answer:Most ad agencies are scared of the technical bar。

Writing a TVC script or building a media plan is not the same skill set as writing Python, running RAG retrieval, or tuning system prompts.The talent structure is different。

Satsuma can pull this off because:

  1. The founder is dual-trained: saomin writes code himself (originally an engineer, later moved into marketing). Sonothing gets outsourced — we deliver in-house
  2. 20 years of game industry experience: games werethe earliest industry to treat "conversational NPCs" as a marketing tool(every game character is a brand ambassador). AI customer service is essentially "extending NPC dialogue to the website" — we've been working with that metaphor for 15 years
  3. Client structure supports it: most of our clients are mid-size brands (budgets NT$50K-200K), which happens to bethe sweet spot for custom AI teammates— too small for SaaS, too unglamorous for big agencies

→ Coincidence + structureput Satsuma in exactly this niche. For other ad agencies to do the same would mean restructuring teams — a 1-2 year project.

Core differences between our AI teammates and SaaS

(We make this argument more fully in this article → . Condensed here.)

Dimension SaaS Satsuma AI Teammate
Character design Generic assistant Character consistent with the TVC / brand universe
Update cadence Client updates it themselves Updated in sync with ad campaigns
Funnel data Not shared with the agency We're the agency — we use it ourselves
Brand tone Template-style pleasantries Voice designed from the brand book
Ad / social integration Each in its own silo One chain

The key is the last row:"One chain"。

When ad campaigns are updated, wealso update Xiao'Ai's KB; when the social campaign copy changes,we also adjust the voice; when we spot a drop-off in the funnel,we also tune the system prompt。

→ We're not selling AI customer service. We're selling "advertising + customer service, one chain."。

What actual clients say

(Below are anonymized, composite representative responses.)

"Our chatbot used to be bought by the marketing team. As the business owner, I'd never seen the backend.Now I get a weekly summary of Xiao'Ai's conversations, and I finally know what customers are actually asking." — CEO, mid-size subscription service

"The people IG ads used to bring inwould bounce 80% of the time after landing. Now Xiao'Ai catches them in the bottom right, asks what they need,and conversion is up 35%." — Marketing lead at a brand

"We got quotes from both you and a SaaS. The SaaS monthly fee was cheaper, but the SaaS didn't give me the feeling that 'this chatbot belongs to my brand.'" — Service industry decision-maker

What it means for you

If you're evaluating AI customer service, ask yourself three questions:

1. Are your site visitors driven by advertising?

  • Yes → A generic SaaS chatbot disconnected from the ad's world will hurt your conversion rate.Find an AI teammate that extends the ad atmosphere
  • No (mostly SEO / direct search) → SaaS may be enough

2. Does your brand have a brand book?

  • Yes → A generic SaaS persona will dilute your brand.Get a custom persona
  • No → Then SaaS works

3. Do you want to optimize the whole funnel, or just add one more tool?

  • The whole funnel → Work with the people doing your advertising on AI customer service
  • Just one more tool → SaaS will do

→ Clients who answered "yes / yes / whole funnel" to all three are a fit for Satsuma. Others aren't.

How to start

  1. Want to try it:Satsuma's homepage Xiao'Ai in the bottom right— feel "an AI that speaks the brand's voice" first-hand
  2. See the full argument:Full AI teammate intro →
  3. 30-minute brief:Fill out the inquiry form, no charge, you'll get a feasibility take back same-day

Further reading: - Stop buying AI customer service SaaS → - The real cost of raising an AI teammate → - AI customer service selection guide →

Satsuma Creative

Integrated marketing creative agency. Creative is the core, technology is the extension — letting AI teammates catch the visitors that need to be caught, at the last mile of the ad funnel.

Related articles · Customer service

Connecting AI customer service to a LINE Official Account: the 4 pitfalls Taiwanese brands most often hit
The hard part of LINE AI customer service isn't the technology, it's the design and the workflow. No entry-point design, old keyword rules clashing, context lost on handoff to a human, push notifications disconnected from the knowledge base — all four pitfalls are people problems.
Intercom, Zendesk, custom AI: how should a mid-size Taiwanese brand choose?
Monthly fees across the three paths range from a few thousand to a few hundred thousand, but the bigger gap is what you're actually buying. The core question for selection is just one: is "ticket management" more important to your customer service, or is "conversation quality"?
FAQ Bot vs. AI Teammate: both auto-reply — what's the difference?
Many brands install "AI customer service" but what they've actually installed is a FAQ Bot. The two look alike, but the logic underneath is completely different. Most brands complaining "AI customer service is dumb" aren't actually using AI at all.