AI Automation
Agentic AI
Enterprise Saas
Multi-Agent Systems
Non-Technical Users
Easy Agent For You
Easy Agent For You
Easy Agent For You
Redesigning an AI agent platform so non-technical users and teams can create, train, and work with agents without writing prompts.
Redesigning an AI agent platform so non-technical users and teams can create, train, and work with agents without writing prompts.
Redesigning an AI agent platform so non-technical users and teams can create, train, and work with agents without writing prompts.
AI Automation
Agentic AI
Agentic AI
Enterprise SaaS
Multi-Agent Systems
Multi-Agent Systems
Non-Technical Users
Non-Technical Users
Role
Role
Lead Product Designer
Lead Product Designer
Lead Product Designer
Scope
Scope
V1 Redesign → V2 B2C + B2B Enterprise Platform
V1 Redesign → V2 B2C + B2B Enterprise Platform
V1 Redesign → V2 B2C + B2B Enterprise Platform
Platform
Platform
Web Application
Web Application
Web Application
Type
Type
Sole Designer, End-to-End
Sole Designer, End-to-End
Sole Designer, End-to-End
I was handed a product that confused its own clients in a live presentation. I redesigned it into a shipped AI platform for individuals and whole organisations, designed so non-technical users can create and use agents without documentation and without knowing what a prompt is.
I was handed a product that confused its own clients in a live presentation. I redesigned it into a shipped AI platform for individuals and whole organisations, designed so non-technical users can create and use agents without documentation and without knowing what a prompt is.
I was handed a product that confused its own clients in a live presentation. I redesigned it into a shipped AI platform for individuals and whole organisations, designed so non-technical users can create and use agents without documentation and without knowing what a prompt is.
The Problem
The Problem
A product that stopped people before they started
A product that stopped people before they started
A product that stopped people before they started
At CISO Middle East Summit 2026 in Doha, the CEO presented Easy Agent V1 to enterprise clients. They compared it unfavourably to Cortech Talent V2, another product I had designed. The technology was capable. The experience asked users to understand AI before they could use it.
V2 also had a bigger job: supporting agents across a whole organisation, with ownership, permissions, sharing, teams, and public deployment, without turning into an infrastructure console.
At CISO Middle East Summit 2026 in Doha, the CEO presented Easy Agent V1 to enterprise clients. They compared it unfavourably to Cortech Talent V2, another product I had designed. The technology was capable. The experience asked users to understand AI before they could use it.
V2 also had a bigger job: supporting agents across a whole organisation, with ownership, permissions, sharing, teams, and public deployment, without turning into an infrastructure console.
Failure
Failure
What was broken
What was broken
Blank Behavioural Prompt
Blank Behavioural Prompt
Blank Behavioural Prompt
No guidance. Users left it empty or wrote vague instructions, producing unpredictable agents.
No guidance. Users left it empty or wrote vague instructions, producing unpredictable agents.
No guidance. Users left it empty or wrote vague instructions, producing unpredictable agents.
Creation Blocked by Training
Creation Blocked by Training
Creation Blocked by Training
Training was collapsed into the creation wizard. Real business data ground the flow to a halt mid-setup.
Training was collapsed into the creation wizard. Real business data ground the flow to a halt mid-setup.
Training was collapsed into the creation wizard. Real business data ground the flow to a halt mid-setup.
Blank New User State
Blank New User State
New users on Ask AI had no agent, no conversation, no visible next action. Zero recovery path.
New users on Ask AI had no agent, no conversation, no visible next action. Zero recovery path.
New users on Ask AI had no agent, no conversation, no visible next action. Zero recovery path.
No Data Boundary
No Data Boundary
No Data Boundary
Sensitive internal data could silently reach public bots. The user never knew the risk existed.
V1: A hectic step without user knowing what he is doing and why
V1: A hectic step without user knowing what he is doing and why
V2: Simplified configuration with a right panel showing behavior prompts based on selections.
V2: Simplified configuration with a right panel showing behavior prompts based on selections.
"V1 asked users to understand AI before using it. V2 gives them five questions they already know how to answer."
"V1 asked users to understand AI before using it. V2 gives them five questions they already know how to answer."
At CISO Middle East Summit 2026 in Doha, the CEO presented Easy Agent V1 to enterprise clients. They compared it unfavourably to Cortech Talent V2, another product I had designed. The technology was capable. The experience asked users to understand AI before they could use it.
V2 also had a bigger job: supporting agents across a whole organisation, with ownership, permissions, sharing, teams, and public deployment, without turning into an infrastructure console.
"V1 asked users to understand AI before using it. V2 gives them five questions they already know how to answer."
V1: A hectic step without user knowing what he is doing and why
V2: Simplified configuration with a right panel showing behavior prompts based on selections.
My Role
My Role
Every decision. One designer.
Every decision. One designer.
Every decision. One designer.
I owned the product design from the V1 audit through the shipped V2: research, architecture, information architecture, UX flows, UI, system states, and developer handoff.
Leadership set the vision: take Easy Agent from individual agents to an enterprise platform. My role was defining how that vision became the shipped product. I researched and designed its structure, workflows, interaction rules, states, and edge cases, then reviewed, challenged, and refined them with leadership and engineering through implementation. Several parts of the final system were reshaped or introduced along the way, rather than coming from the original plan.
I owned the product design from the V1 audit through the shipped V2: research, architecture, information architecture, UX flows, UI, system states, and developer handoff.
Leadership set the vision: take Easy Agent from individual agents to an enterprise platform. My role was defining how that vision became the shipped product. I researched and designed its structure, workflows, interaction rules, states, and edge cases, then reviewed, challenged, and refined them with leadership and engineering through implementation. Several parts of the final system were reshaped or introduced along the way, rather than coming from the original plan.
The CEO's brief was direct: a complete version, with no loopholes, no dead ends, and no dropped use cases. Where something was undefined, I had the freedom to define it.
The CEO's brief was direct: a complete version, with no loopholes, no dead ends, and no dropped use cases. Where something was undefined, I had the freedom to define it.
The CEO's brief was direct: a complete version, with no loopholes, no dead ends, and no dropped use cases. Where something was undefined, I had the freedom to define it.
To close the gap between design and development, I created a handoff format the team did not have before: 12 module-level documents, each with SQL models, API contracts, ASCII wireframes, UI states, edge cases, and acceptance criteria.
To close the gap between design and development, I created a handoff format the team did not have before: 12 module-level documents, each with SQL models, API contracts, ASCII wireframes, UI states, edge cases, and acceptance criteria.
To close the gap between design and development, I created a handoff format the team did not have before: 12 module-level documents, each with SQL models, API contracts, ASCII wireframes, UI states, edge cases, and acceptance criteria.
I owned the product design from the V1 audit through the shipped V2: research, architecture, information architecture, UX flows, UI, system states, and developer handoff.
Leadership set the vision: take Easy Agent from individual agents to an enterprise platform. My role was defining how that vision became the shipped product. I researched and designed its structure, workflows, interaction rules, states, and edge cases, then reviewed, challenged, and refined them with leadership and engineering through implementation. Several parts of the final system were reshaped or introduced along the way, rather than coming from the original plan.
System Architecture
System Architecture
12 modules, one system
12 modules, one system
12 modules, one system
V2 spanned 12 interconnected modules, from agent creation and training to automation, public deployment, permissions, Agent Teams, and analytics. An agent created in one place could be trained elsewhere, assigned to someone, placed in a team, or deployed publicly, so no module could be designed alone. I mapped the product as a system of relationships and states first, across four FigJam workflow diagrams, and resolved contradictions between modules at wireframe stage. The screens came afterwards.
V2 spanned 12 interconnected modules, from agent creation and training to automation, public deployment, permissions, Agent Teams, and analytics. An agent created in one place could be trained elsewhere, assigned to someone, placed in a team, or deployed publicly, so no module could be designed alone. I mapped the product as a system of relationships and states first, across four FigJam workflow diagrams, and resolved contradictions between modules at wireframe stage. The screens came afterwards.
To close the gap between design and development, I created a handoff format the team did not have before: 12 module-level documents, each with SQL models, API contracts, ASCII wireframes, UI states, edge cases, and acceptance criteria.
I owned the product design from the V1 audit through the shipped V2: research, architecture, information architecture, UX flows, UI, system states, and developer handoff.
Leadership set the vision: take Easy Agent from individual agents to an enterprise platform. My role was defining how that vision became the shipped product. I researched and designed its structure, workflows, interaction rules, states, and edge cases, then reviewed, challenged, and refined them with leadership and engineering through implementation. Several parts of the final system were reshaped or introduced along the way, rather than coming from the original plan.
Research and Discovery
Research and Discovery
The finding that changed everything
The finding that changed everything
The finding that changed everything
The Doha feedback gave me users who could already name exactly where they got lost. I audited V1 with heuristic evaluation and cognitive walkthroughs against one filter: can a non-technical business owner complete this without help? The answer was consistently no.
I studied ChatGPT, Zapier, and Intercom to understand familiar mental models around conversation, automation, and business support. The opportunity for Easy Agent was to make agent creation approachable for non-technical users, without giving up control over data and access.
The Doha feedback gave me users who could already name exactly where they got lost. I audited V1 with heuristic evaluation and cognitive walkthroughs against one filter: can a non-technical business owner complete this without help? The answer was consistently no.
I studied ChatGPT, Zapier, and Intercom to understand familiar mental models around conversation, automation, and business support. The opportunity for Easy Agent was to make agent creation approachable for non-technical users, without giving up control over data and access.
Core finding: The setup came before the understanding. Users were asked to configure an agent before they had any context for what it would do. They did not lack motivation; they lacked a mental model. The problem was never the technology. It was the missing bridge.
Core finding: The setup came before the understanding. Users were asked to configure an agent before they had any context for what it would do. They did not lack motivation; they lacked a mental model. The problem was never the technology. It was the missing bridge.
Core finding: The setup came before the understanding. Users were asked to configure an agent before they had any context for what it would do. They did not lack motivation; they lacked a mental model. The problem was never the technology. It was the missing bridge.
The Doha client feedback gave me something unusual: users who could already name exactly where they got lost. Usually you have to observe confusion. Here it had been surfaced in a live presentation and stated directly.
I ran V1 through a heuristic audit with one filter: can a non-technical business owner complete this without external help? The answer was consistently no.
Competitive analysis across ChatGPT, Zapier, and Intercom confirmed the same pattern. Every powerful tool demanded technical literacy as the price of entry.
Product
Product
Product
ChatGPT
ChatGPT
ChatGPT
Zapier
Zapier
Zapier
Intercom
Intercom
Intercom
Easy Agent V1
Easy Agent V1
Easy Agent V1
Easy Agent V2 (only one)
Easy Agent V2 (only one)
Easy Agent V2 (only one)
Requires prompt writing
Requires prompt
writing
Requires prompt writing
Requires technical setup
Requires
technical setup
Requires technical setup
Guided creation flow
Guided
creation flow
Guided creation flow
Non-technical first
Non-technical
first
Non-technical first
Core user perspectives
Core user perspectives
Core user perspectives
Non-Technical Business Owner
Non-Technical Business Owner
Knows what they want the agent to do. Has no idea what a system prompt is.
Knows what they want the agent to do. Has no idea what a system prompt is.
Support Team Lead
Support Team Lead
Needs an agent that answers consistently from controlled business knowledge.
Needs an agent that answers consistently from controlled business knowledge.
Needs an agent that answers consistently from controlled business knowledge.
Operations Manager
Operations Manager
Wants agents connected to tools and people, without losing oversight.
Wants agents connected to tools and people, without losing oversight.
Enterprise Owner / Admin
Enterprise Owner / Admin
Owns the enterprise subscription and distributes agents and quota across teams and employees. Needs clear control over creation, assignment, permissions, and destructive actions.
Owns the enterprise subscription and distributes agents and quota across teams and employees. Needs clear control over creation, assignment, permissions, and destructive actions.
Key Design Decisions
Key Design Decisions
The seven decisions that rebuilt the product
The seven decisions that rebuilt the product
The seven decisions that rebuilt the product
Decision 01 — Most Strategic
Decision 01 — Most Strategic
The blank prompt becomes five questions
The blank prompt becomes five questions
The blank prompt becomes five questions
The single biggest drop-off point in V1 was a text field nobody knew how to fill.
The single biggest drop-off point in V1 was a text field nobody knew how to fill.
The single biggest drop-off point in V1 was a text field nobody knew how to fill.
The "Behavioural Prompt" textarea placed the hardest possible task first and gave users nothing to work with. I replaced it with five plain-language questions covering agent type, communication style, primary goal, response depth, and domain focus.
As the user answers the generated prompt updates live in a panel on the right. Users watch their answers turn into a working prompt in real time without writing a single word themselves.
The barrier to first value dropped from "understand what a system prompt is and write a good one" to "answer five questions you already know the answers to."
The "Behavioural Prompt" field placed the hardest possible task first and gave users nothing to work with. I replaced it with five plain-language questions covering agent type, communication style, primary goal, response depth, and domain focus.
As the user answers the generated prompt updates live in a panel on the right. Users watch their answers turn into a working prompt in real time without writing a single word themselves.
The "Behavioural Prompt" field placed the hardest possible task first and gave users nothing to work with. I replaced it with five plain-language questions covering agent type, communication style, primary goal, response depth, and domain focus.
As the user answers the generated prompt updates live in a panel on the right. Users watch their answers turn into a working prompt in real time without writing a single word themselves.
"V1 asked users to understand AI before using it. V2 starts with five questions they already know how to answer."
"V1 asked users to understand AI before using it. V2 starts with five questions they already know how to answer."
"The hardest step in V1 was the first one. V2 makes the first step the easiest one."
Decision 02
Decision 02
Creation and training become two separate moments
Creation and training become two separate moments
Creation and training become two separate moments
Knowing what you want an agent to do and having your data ready are two different moments.
Knowing what you want an agent to do and having your data ready are two different moments.
Knowing what you want an agent to do and having your data ready are two different moments.
The original wizard tried to define the agent and train it in the same flow. For businesses with multiple apps and live APIs, this created a flow that stopped dead mid-setup.
I split these completely. Creation became a three step flow under two minutes. Training became its own dedicated page reachable any time after creation.
A user can have a working agent in under two minutes without uploading a single file.
The original wizard tried to define the agent and train it in the same flow. For businesses with multiple apps and live APIs, that flow stopped dead mid-setup.
I split them completely. Creation sets the agent's identity, role, and behaviour in a three-step flow. Training, which covers both knowledge and connected tools such as MCP connections and APIs, became its own page, reachable any time after creation.
"A user can have a working agent in under two minutes without uploading a single file."
The original wizard tried to define the agent and train it in the same flow. For businesses with multiple apps and live APIs, that flow stopped dead mid-setup.
I split them completely. Creation sets the agent's identity, role, and behaviour in a three-step flow. Training, which covers both knowledge and connected tools such as MCP connections and APIs, became its own page, reachable any time after creation.
"A user can have a working agent in under two minutes without uploading a single file."
"Separating creation from training removed the biggest setup blocker without removing any capability."
"Separating creation from training removed the biggest setup blocker without removing any capability."
"Separating creation from training removed the biggest setup blocker without removing any capability."
Decision 03
Decision 03
Aria turns a technical fallback into a first-use experience
Aria turns a technical fallback into a first-use experience
Aria turns a technical fallback into a first-use experience
A new user's first message should never go nowhere.
A new user's first message should never go nowhere.
A new user's first message should never go nowhere.
Every new user in V1 landed on Ask AI to find nothing no agent no conversation no visible next action. The most common behaviour was inaction. They waited for something to happen. Nothing did. That was the drop off point.
I introduced Aria, a pre-built agent trained on Easy Agent's own documentation, that activates the moment a new user sends their first message.
Aria is intentionally locked at the system level. Users cannot rename retrain or delete it. This was a deliberate product decision if Aria could be misconfigured the onboarding safety net disappears for every user who comes after.
In V1, the most common behaviour of a new user was inaction: they waited for something to happen, and nothing did. Building on an initial technical solution from our CTO, a system agent that let the first conversation work, I designed the product experience it became: Aria.
Aria is a pre-trained agent built on Easy Agent's own documentation, and it activates the moment a new user sends their first message. It is locked at the system level: users cannot rename, retrain, or delete it. If Aria could be misconfigured, the onboarding safety net would disappear for every user who came after.
In V1, the most common behaviour of a new user was inaction: they waited for something to happen, and nothing did. Building on an initial technical solution from our CTO, a system agent that let the first conversation work, I designed the product experience it became: Aria.
Aria is a pre-trained agent built on Easy Agent's own documentation, and it activates the moment a new user sends their first message. It is locked at the system level: users cannot rename, retrain, or delete it. If Aria could be misconfigured, the onboarding safety net would disappear for every user who came after.
"Every new user has somewhere to go from the first message."
"Every new user has somewhere to go from the first message."
"Every new user has somewhere to go from the first message."
Decision 04 · Most Protective
Decision 04 · Most Protective
The public bot never directly exposes the private agent
The public bot never directly exposes the private agent
The public bot never directly exposes the private agent
Safety that depends on the user configuring it correctly is not safety.
Safety that depends on the user configuring it correctly is not safety.
Safety that depends on the user configuring it correctly is not safety.
When a user publishes an agent as a customer facing bot the naive default gives that bot access to everything the parent agent knows including private Slack channels and confidential documents.
I designed Step 2 of the Bot Studio wizard so nothing is selected by default. The user explicitly ticks what the bot is allowed to see. The language is human: "choose what this bot is allowed to know."
The safe option is the default. Protection does not require the user to understand the risk.
When a user publishes an agent as a public bot, the naive default gives that bot access to everything the parent agent knows, including private Slack channels and confidential documents. A settings screen alone would still leave protection up to the user, so I proposed a separate Deployment Agent layer between the source agent and the public bot.
At deployment, the user publishes the original agent or creates a scoped sub-agent. In Step 2 of the Bot Studio wizard, nothing is selected by default: the user explicitly chooses what the bot may know, in plain language: "choose what this bot is allowed to know." A scoped version also gets its own live MCP connections, APIs, and behavioural instructions.
The parent agent is the brain. The Deployment Agent controls what a public bot can access and use.
When a user publishes an agent as a public bot, the naive default gives that bot access to everything the parent agent knows, including private Slack channels and confidential documents. A settings screen alone would still leave protection up to the user, so I proposed a separate Deployment Agent layer between the source agent and the public bot.
At deployment, the user publishes the original agent or creates a scoped sub-agent. In Step 2 of the Bot Studio wizard, nothing is selected by default: the user explicitly chooses what the bot may know, in plain language: "choose what this bot is allowed to know." A scoped version also gets its own live MCP connections, APIs, and behavioural instructions.
The parent agent is the brain. The Deployment Agent controls what a public bot can access and use.
"Nothing is shared until the user explicitly chooses it. Safe is the default, not the setting."
"Nothing is shared until the user explicitly chooses it. Safe is the default, not the setting."
"Nothing is shared until the user explicitly chooses it. Safe is the default, not the setting."
Decision 05 · Most Architecturally Complex
Decision 05 · Most Architecturally Complex
From individual assistants to an organisational AI system
From individual assistants to an organisational AI system
From individual assistants to an organisational AI system
Individual assistants alone would recreate a problem every company already has: knowledge stuck with individuals.
Individual assistants alone would recreate a problem every company already has: knowledge stuck with individuals.
Individual assistants alone would recreate a problem every company already has: knowledge stuck with individuals.
The sidebar needed to hold ten modules. Showing all ten at equal visual weight on a first visit is not an information architecture problem it is a retention problem. The product looks more complicated than it is before the user has done anything.
I grouped Integration, Ticket Management, and Payments into a collapsed "More" section at the bottom. The six primary modules stay permanently visible. Secondary modules are one click away but they do not compete for attention on a first visit.
The enterprise vision was an AI assistant for every employee. But real organisations delegate, share across departments, and hand over access when someone is busy. So I designed agents around the organisation: a CEO can own a high-level agent, and an MD can request limited, constrained access to it.
That foundation became Agent Team, our user-facing Multi-Agent System. I researched multi-agent systems and translated the architecture into a model every manager already understands: a team. The orchestrator is the Team Leader, specialist agents are the Team Members, and the user simply talks to the team.
The enterprise vision was an AI assistant for every employee. But real organisations delegate, share across departments, and hand over access when someone is busy. So I designed agents around the organisation: a CEO can own a high-level agent, and an MD can request limited, constrained access to it.
That foundation became Agent Team, our user-facing Multi-Agent System. I researched multi-agent systems and translated the architecture into a model every manager already understands: a team. The orchestrator is the Team Leader, specialist agents are the Team Members, and the user simply talks to the team.
"The collapse is not about aesthetics. It is about the three seconds between a first visit and a decision to stay."
"The collapse is not about aesthetics. It is about the three seconds between a first visit and a decision to stay."
"The collapse is not about aesthetics. It is about the three seconds between a first visit and a decision to stay."
Decision 06 — Most Operationally Distinctive
Decision 06 — Most Operationally Distinctive
Decision 06 — Most Operationally Distinctive
The Asset Store as a two-sided marketplace with distinct product types
The Asset Store as a two-sided marketplace with distinct product types
The Asset Store as a two-sided marketplace with distinct product types
V1 forced every user to start from zero. The Asset Store gave non-technical users a starting point that already worked.
V1 forced every user to start from zero. The Asset Store gave non-technical users a starting point that already worked.
V1 forced every user to start from zero. The Asset Store gave non-technical users a starting point that already worked.
I introduced the Asset Store with two distinct product types. Ready Agents are pre-trained agents that install into the Agents module. UI Templates are output format structures that install into Automation so results arrive in a structured layout rather than plain text.
Blurring the two product types would have broken the user's mental model of how the system works. An agent is something you train and converse with. A template is a structure that your automation outputs into. Treating them as interchangeable in a marketplace would have made both feel unreliable.
The product shifted from a blank canvas that demanded expertise into a platform where a non-technical user can start from something that already works.
I introduced the Asset Store with two distinct product types. Ready Agents are pre-trained agents that install into the Agents module. UI Templates are output format structures that install into Automation so results arrive in a structured layout rather than plain text.
Blurring the two product types would have broken the user's mental model of how the system works. An agent is something you train and converse with. A template is a structure that your automation outputs into. Treating them as interchangeable in a marketplace would have made both feel unreliable.
The product shifted from a blank canvas that demanded expertise into a platform where a non-technical user can start from something that already works.
I introduced the Asset Store with two distinct product types. Ready Agents are pre-trained agents that install into the Agents module. UI Templates are output format structures that install into Automation so results arrive in a structured layout rather than plain text.
Blurring the two product types would have broken the user's mental model of how the system works. An agent is something you train and converse with. A template is a structure that your automation outputs into. Treating them as interchangeable in a marketplace would have made both feel unreliable.
The product shifted from a blank canvas that demanded expertise into a platform where a non-technical user can start from something that already works.
"Ready Agents install into your Agents module. Templates install into Automation. Different products. Different destinations."
"Ready Agents install into your Agents module. Templates install into Automation. Different products. Different destinations."
"Ready Agents install into your Agents module. Templates install into Automation. Different products. Different destinations."
The Design Principles I Proved
The Design Principles I Proved
Every principle backed by a specific decision
Every principle backed by a specific decision
Every principle backed by a specific decision
Complexity hidden until needed
→ Proved in Decision 01 and 05
The behavioural prompt is generated, not written. Secondary navigation is collapsed, not removed.
Guided before open
→ Proved in Decision 01, 02, and 03
Every technically complex moment has a guided alternative. Aria guides before the user understands the product.
Safe by default
→ Proved in Decision 04
The data boundary defaults to nothing shared. Protection requires zero user understanding of the risk.
First impressions are designed, not accidental
→ Proved in Decision 03
Aria eliminates the blank state. Every new user has a conversation to start from the first message.
Complexity hidden until needed
→ Proved in Decision 01 and 05
→ Proved in Decision 01 and 05
The behavioural prompt is generated, not written. Secondary navigation is collapsed, not removed.
The behavioural prompt is generated, not written. Secondary navigation is collapsed, not removed.
Guided before open
→ Proved in Decision 01, 02, and 03
→ Proved in Decision 01, 02, and 03
Every technically complex moment has a guided alternative. Aria guides before the user understands the product.
Every technically complex moment has a guided alternative. Aria guides before the user understands the product.
Safe by default
→ Proved in Decision 04
→ Proved in Decision 04
The data boundary defaults to nothing shared. Protection requires zero user understanding of the risk.
The data boundary defaults to nothing shared. Protection requires zero user understanding of the risk.
First impressions are designed, not accidental
→ Proved in Decision 03
→ Proved in Decision 03
Aria eliminates the blank state. Every new user has a conversation to start from the first message.
Aria eliminates the blank state. Every new user has a conversation to start from the first message.
Outcome
Outcome
What the product became
What the product became
What the product became
Easy Agent For You V1 was failing at live enterprise demos with near-zero user traction. V2 redesign took the platform to production and shipped to enterprise clients. Website search discovery grew from near-zero to 471 clicks and 113,000 impressions over 3 months, position 17.2, CTR 0.4%. CEO confirmed subscription-based user adoption increased. 12 modules redesigned end-to-end. 0 blank states remaining. Every module shipped and in active use.
Easy Agent For You V1 was failing at live enterprise demos with near-zero user traction. V2 redesign took the platform to production and shipped to enterprise clients. Website search discovery grew from near-zero to 471 clicks and 113,000 impressions over 3 months, position 17.2, CTR 0.4%. CEO confirmed subscription-based user adoption increased. 12 modules redesigned end-to-end. 0 blank states remaining. Every module shipped and in active use.
12
12
Modules redesigned end to end from agent creation through enterprise portal each with zero dropped use cases
Modules redesigned end to end from agent creation through enterprise portal each with zero dropped use cases
0
0
Blank states remaining every empty error and edge case has a defined next action
Blank states remaining every empty error and edge case has a defined next action
V1 → V2
V1 → V2
The version that was failing at live enterprise demos became the version shipped to enterprise clients. Website search discovery grew from near-zero to 471 clicks and 113,000 impressions over 3 months, position 17.2, CTR 0.4%. Upward trajectory confirmed. CEO separately confirmed subscription-based user adoption increased.
The version that was failing at live enterprise demos became the version shipped to enterprise clients. Website search discovery grew from near-zero to 471 clicks and 113,000 impressions over 3 months, position 17.2, CTR 0.4%. Upward trajectory confirmed. CEO separately confirmed subscription-based user adoption increased.
12
Modules redesigned end to end from agent creation through enterprise portal each with zero dropped use cases
0
Blank states remaining every empty error and edge case has a defined next action
V1 → V2
The version that was failing at live enterprise demos became the version shipped to enterprise clients. Website search discovery grew from near-zero to 471 clicks and 113,000 impressions over 3 months, position 17.2, CTR 0.4%. Upward trajectory confirmed. CEO separately confirmed subscription-based user adoption increased.
Easy Agent For You V1 was failing at live enterprise demos with near-zero user traction. V2 redesign took the platform to production and shipped to enterprise clients. Website search discovery grew from near-zero to 471 clicks and 113,000 impressions over 3 months, position 17.2, CTR 0.4%. CEO confirmed subscription-based user adoption increased. 12 modules redesigned end-to-end. 0 blank states remaining. Every module shipped and in active use.
The B2B model of Easy Agent is a very grand setup almost in comparison to any large AI model like Anthropic Claude working in the market. The current design reflects a depth of understanding with more focus on user usability and great improvements in flows.
The B2B model of Easy Agent is a very grand setup almost in comparison to any large AI model like Anthropic Claude working in the market. The current design reflects a depth of understanding with more focus on user usability and great improvements in flows.
The B2B model of Easy Agent is a very grand setup almost in comparison to any large AI model like Anthropic Claude working in the market. The current design reflects a depth of understanding with more focus on user usability and great improvements in flows.
— CEO, Cortechsols
— CEO, Cortechsols
Reflection
Reflection
The hardest part of designing AI products is not the AI. It is the gap between what the system is doing and what the user believes they are doing. Easy Agent V2 is more powerful than V1 not because more features were added but because the friction between the user and the existing features was removed. That is almost always where the real design work is.
The hardest part of designing AI products is not the AI. It is the gap between what the system is doing and what the user believes they are doing. Easy Agent V2 is more powerful than V1 not because more features were added but because the friction between the user and the existing features was removed. That is almost always where the real design work is.
The hardest part of designing AI products is not the AI. It is the gap between what the system is doing and what the user believes they are doing. Easy Agent V2 is more powerful than V1 not because more features were added but because the friction between the user and the existing features was removed. That is almost always where the real design work is.