Behavioral Design | AI Experience Design | Product & Design Strategy
Behavioral Design | AI Experience Design | Product & Design Strategy
AI Call Agent for SMBs drowning in customer calls
AI Call Agent for SMBs drowning in customer calls
A design built to earn the trust of owners handing their customers to AI for the first time
A design built to earn the trust of owners handing their customers to AI for the first time
SUPERFONE
SUPERFONE
Lead Designer
Lead Designer
2025
2025
04 weeks
04 weeks
0 → 1 feature
0 → 1 feature
mobile
mobile
android
android
live in prod
live in prod
Context
Context
Superfone gives small Indian businesses a virtual phone number their whole team can share, with multiple calls running at once. For most of these businesses (manufacturing, travel agents, design studios) the phone is the business. A ringing phone is a customer. A missed one is a lost customer.
The AI Agent answers the calls the team can't: after business hours, when no one is free, or when call volume outstrips the lines available. Over time it also frees staff from repetitive questions so they can focus on the customers in front of them.
Superfone gives small Indian businesses a virtual phone number their whole team can share, with multiple calls running at once. For most of these businesses (manufacturing, travel agents, design studios) the phone is the business. A ringing phone is a customer. A missed one is a lost customer.
The AI Agent answers the calls the team can’t: after business hours, when no one is free, or when call volume outstrips the lines available. Over time it also frees staff from repetitive questions so they can focus on the customers in front of them.
What it promises the business:
Never miss a call. Agent always answers when the team is busy or closed.
Works 24×7. The business stays reachable round the clock.
Always professional. Sounds like a trained receptionist for that business.
Handles the request. Asks the right questions or shares common information.
Frees up the team. Staff focus on in-store customers and higher-value work.
What it promises the business:
Never miss a call. Agent always answers when the team is busy or closed.
Works 24×7. The business stays reachable round the clock.
Always professional. Sounds like a trained receptionist for that business.
Handles the request. Asks the right questions or shares common information.
Frees up the team. Staff focus on in-store customers and higher-value work.
Lifecycle of a customer call
Lifecycle of a customer call
When an incoming call rings for the team, if someone answers, the call is handled and logged, and its context is shared with the entire team. If no one answers, the call ends, and the customer receives a missed call message while the call log records the missed attempt. This unanswered scenario represents the dead end that this feature aims to eliminate.
When an incoming call rings for the team, if someone answers, the call is handled and logged, and its context is shared with the entire team. If no one answers, the call ends, and the customer receives a missed call message while the call log records the missed attempt. This unanswered scenario represents the dead end that this feature aims to eliminate.
Swipe horizontally
Where the AI agent fits
Where the AI agent fits
Today, an incoming call rings the team. If someone picks up, the call is handled, logged, and its context is shared with the whole team. If no one answers, the branch ends: the customer gets a missed-call message, and the call log records the miss. That unanswered branch is the dead end this feature exists to close.
Today, an incoming call rings the team. If someone picks up, the call is handled, logged, and its context is shared with the whole team. If no one answers, the branch ends: the customer gets a missed-call message, and the call log records the miss. That unanswered branch is the dead end this feature exists to close.
Swipe horizontally
The problem
The problem
The users are not early adopters. Most have never set up an automation of any kind and are unsure of AI. They are cautious about switching it on and need hand-holding.
The users are not early adopters. Most have never set up an automation of any kind and are unsure of AI. They are cautious about switching it on and need hand-holding.
The agent is autonomous. No human is involved during the conversation. Owners are hesitant to hand something as important as a customer call to AI that might get it wrong.
The agent is autonomous. No human is involved during the conversation. Owners are hesitant to hand something as important as a customer call to AI that might get it wrong.
A bad call costs more than a missed one. A poorly trained agent can annoy a caller and lose a lead; a poor setup can drop important calls. The stakes sit on the business, not the tool.
A bad call costs more than a missed one. A poorly trained agent can annoy a caller and lose a lead; a poor setup can drop important calls. The stakes sit on the business, not the tool.
Design Principles
Design Principles
1
1
Build trust
Build trust
Owners will only trust what they can monitor. A poorly configured agent can frustrate callers and lead to missed calls, harming the business. Therefore, the feature must instil confidence that the agent will take care of customers, allowing owners to focus on other tasks. This confidence comes from giving owners real control over the agent's behaviour, ensuring transparency during calls, and a simple setup process. When owners can see and modify what the agent will say and review calls afterward, they gain tangible control over the situation.
Owners will only trust what they can monitor. A poorly configured agent can frustrate callers and lead to missed calls, harming the business. Therefore, the feature must instil confidence that the agent will take care of customers, allowing owners to focus on other tasks. This confidence comes from giving owners real control over the agent’s behaviour, ensuring transparency during calls, and a simple setup process. When owners can see and modify what the agent will say and review calls afterward, they gain tangible control over the situation.
2
2
Agent as a team member
Agent as a team member
The agent is more than just a tool; it's like a dedicated coworker available 24/7, handling conversations just like a trained staff member. This perspective reinforces the owner’s responsibility for the agent's setup, likening it to onboarding a real employee by adding knowledge and defining roles.
The agent's persona is also vital. By choosing its voice, name, and personality, the owner is selecting the representative for the business. This makes the agent a part of the team, featured prominently alongside human members, reflecting its importance and the care it deserves.
The agent is more than just a tool; it’s like a dedicated coworker available 24/7, handling conversations just like a trained staff member. This perspective reinforces the owner’s responsibility for the agent’s setup, likening it to onboarding a real employee by adding knowledge and defining roles.
The agent’s persona is also vital. By choosing its voice, name, and personality, the owner is selecting the representative for the business. This makes the agent a part of the team, featured prominently alongside human members, reflecting its importance and the care it deserves.
3
3
Full control and transparency
Full control and transparency
The system uses clear and simple language, avoiding jargon, to outline what will happen during interactions: what the agent will say, when they will respond, and what actions they can and cannot take. The most critical moments occur when the owner cannot observe when the agent is alone with the customer. Therefore, every call is documented with a summary, tags, suggested actions, and a task assigned to the appropriate team member. Control is maintained over the call rather than during it. There is full visibility of the agent's involvement in the call logs to assure owners that the agent is active, and team members can quickly access the information needed to take action.
The system uses clear and simple language, avoiding jargon, to outline what will happen during interactions: what the agent will say, when they will respond, and what actions they can and cannot take. The most critical moments occur when the owner cannot observe when the agent is alone with the customer. Therefore, every call is documented with a summary, tags, suggested actions, and a task assigned to the appropriate team member. Control is maintained over the call rather than during it. There is full visibility of the agent’s involvement in the call logs to assure owners that the agent is active, and team members can quickly access the information needed to take action.
Design strategy
Design strategy
The setup is designed as a guided journey with two main objectives: to earn the owner's trust before the agent handles a real call, and to ensure the owner remains in control afterward. Each step is framed in the owner's terms such as the business context, the types of questions callers may ask, and when the phone should be answered rather than using technical jargon from the underlying system.
The following journey breakdown outlines each stage, detailing what the owner does, the intent behind each design decision, and the technology that facilitates it.
The setup is designed as a guided journey with two main objectives: to earn the owner’s trust before the agent handles a real call, and to ensure the owner remains in control afterward. Each step is framed in the owner’s terms such as the business context, the types of questions callers may ask, and when the phone should be answered rather than using technical jargon from the underlying system.
The following journey breakdown outlines each stage, detailing what the owner does, the intent behind each design decision, and the technology that facilitates it.
Swipe horizontally
Drag horizontally to explore
Swipe horizontally
Swipe horizontally
Call flow complexity
Call flow complexity
Superfone's call routing was already dense before any AI was added. A single incoming call could be shaped by several rules at once:
Business hours — open versus closed, each needing different behaviour.
Ringing order — which team member rings first, and in what groups.
Sticky agent — a call routed directly to the lead's owner.
DND — a team member who has muted direct calls.
IVR — a menu already routing the caller.
Known vs. unknown numbers — existing contacts versus first-time callers.
No answer, busy, or declined — the fallback conditions that pass a call along.
Superfone’s call routing was already dense before any AI was added. A single incoming call could be shaped by several rules at once:
Business hours — open versus closed, each needing different behaviour.
Ringing order — which team member rings first, and in what groups.
Sticky agent — a call routed directly to the lead’s owner.
DND — a team member who has muted direct calls.
IVR — a menu already routing the caller.
Known vs. unknown numbers — existing contacts versus first-time callers.
No answer, busy, or declined — the fallback conditions that pass a call along.
Introducing an autonomous agent raises an important question: when exactly does the agent respond? Each possible answer leads to a different configuration. If we present these as settings, it would be overwhelming for non-technical users, making the setup itself a reason for the feature to fail.
To simplify this, instead of creating a new configuration interface, we can integrate the agent into the existing ringing order as an additional team member. This approach means that determining when the agent answers becomes a matter of seating order, allowing the agent to follow the existing routing rules rather than complicating them. This way, there’s no need for a new mental model or an additional settings screen.
Introducing an autonomous agent raises an important question: when exactly does the agent respond? Each possible answer leads to a different configuration. If we present these as settings, it would be overwhelming for non-technical users, making the setup itself a reason for the feature to fail.
To simplify this, instead of creating a new configuration interface, we can integrate the agent into the existing ringing order as an additional team member. This approach means that determining when the agent answers becomes a matter of seating order, allowing the agent to follow the existing routing rules rather than complicating them. This way, there’s no need for a new mental model or an additional settings screen.
Swipe horizontally
Reflection
Reflection
The most challenging aspect of this project was building trust rather than the technology itself. While creating an agent to answer calls was straightforward, convincing cautious business owners who often unfamiliar with AI and to let the agent interact with their customers was the real hurdle.
Another key lesson was the importance of restraint. Instead of claiming the agent could handle any request, we defined its capabilities realistically. This honesty about its limitations built trust, as providing a confident yet incorrect answer was more harmful than simply missing a call.
The most challenging aspect of this project was building trust rather than the technology itself. While creating an agent to answer calls was straightforward, convincing cautious business owners who often unfamiliar with AI and to let the agent interact with their customers was the real hurdle.
Another key lesson was the importance of restraint. Instead of claiming the agent could handle any request, we defined its capabilities realistically. This honesty about its limitations built trust, as providing a confident yet incorrect answer was more harmful than simply missing a call.
Role and team
Role and team
Sole product designer. Owned end-to-end design from problem framing through final QA. Worked closely with the PM on product strategy and direction, influenced the reframing of the problem and the key design decisions.
Took an active role in QA, including adversarial testing of AI outputs, and partnered with engineering through implementation.
Sole product designer. Owned end-to-end design from problem framing through final QA. Worked closely with the PM on product strategy and direction, influenced the reframing of the problem and the key design decisions.
Took an active role in QA, including adversarial testing of AI outputs, and partnered with engineering through implementation.
Team 1 PM · 2 QA · 1 frontend engineer · 1 backend engineer
Team 1 PM · 2 QA · 1 frontend engineer · 1 backend engineer