Back to work

I designed an AI-powered Assistant Manager to personalize the Fantasy Premier League experience for over 12 million players.

Illustration of a soccer ball, whistle, and tactics clipboard
Role
Product Designer
Project type
Existing Product Iteration Design
Team
Individual Project
Platform
Web & Mobile
Duration
4 Weeks

About the product

Fantasy Premier League requires managers to make weekly decisions based on player form, fixtures, injuries, transfers, expected minutes, and chip strategy. While all of this information is publicly available, interpreting it requires considerable research and experience.

I designed Assistant Manager, an optional premium feature that provides personalized recommendations based on both real-world football data and each user's individual management style. Rather than making decisions for users, the feature explains its reasoning and allows managers to remain in complete control of their squads.

Key observations

There is already demand for expert advice

Millions of Fantasy Premier League managers consume content from YouTube creators, podcasts, Reddit, and statistics platforms before making weekly decisions. This demonstrates that players actively seek guidance, but currently need to gather and interpret information from multiple external sources before finalizing their team.

12M+ players

Fantasy Premier League has over 12 million active managers making weekly decisions across a 38-gameweek season.

No two styles alike

Some managers minimize risk with consistent performers, while others chase high-risk differentials, advice platforms typically give everyone the same answer.

Managers have different philosophies

There is no universally "correct" way to play Fantasy Premier League. Some managers prioritize minimizing risk with consistent performers, while others chase high-risk differential picks. Some remain loyal to players from the club they support, while others focus purely on maximizing points. Existing advice platforms typically provide identical recommendations to everyone, overlooking these individual preferences.

The experience lacks personalization

Despite the strategic nature of Fantasy Premier League, every manager interacts with essentially the same product experience. Aside from differences in squad composition, the platform offers little adaptation to individual playing styles. This creates an opportunity for AI to deliver recommendations that reflect each user's unique management philosophy without changing the core mechanics of the game.

Goals

After evaluating the existing experience, I established the following design goals:

  1. 01

    Reduce the amount of research required before each gameweek.

  2. 02

    Deliver recommendations tailored to each manager's preferred playing style.

  3. 03

    Explain every recommendation transparently rather than acting as a "black box."

  4. 04

    Integrate naturally into the existing FPL interface without disrupting familiar workflows.

Exploring solutions

Option 1

Static weekly advice

A weekly report containing generic recommendations for transfers, captain selections, and lineup changes based solely on current player data.

Why it wasn't chosen: easy to implement but gives every user the same answer, ignoring squad composition, chips, budgets, and playing style, limiting its value over existing creators.

Option 2

Conversational AI only

A chatbot that lets managers ask questions about players, fixtures, transfers, and strategy on demand.

Why it wasn't chosen: flexible, but relies on users knowing what to ask. Less experienced managers don't know which decisions deserve attention, the feature becomes reactive rather than proactively guiding better decisions.

Final direction

Proactive report + conversational follow-up

A personalized weekly report that surfaces the decisions worth attention, paired with a conversational assistant for follow-up, every recommendation shaped by the manager's stated playing style and paired with a confidence rating to communicate uncertainty transparently.

Proposed design

The Assistant Manager lives inside the existing Team Management page that FPL managers already use to pick a squad, make transfers, and set captains. Only onboarding gets its own screen. The weekly report and chatbot are docked into the page where the Player Selection panel normally sits, so nothing about the familiar workflow moves.

Existing FPL Transfers page: pitch view on the left, Player Selection panel on the right.Existing FPL Transfers page mid-transfer, with the Player Selection panel showing search results.
The starting point: FPL's current Team Management / Transfers page. The Assistant Manager slots into the right-hand column beside the pitch, replacing or augmenting the Player Selection panel depending on the mode.

01

Meet your Assistant Manager

Onboarding is the one part of the feature that gets its own screen, shown once during team creation for users who opt into the premium tier. Instead of fixed preferences, every question is answered on a sliding scale, so the system captures each manager's playing style rather than a checklist. These preferences are remembered throughout the season and shape every recommendation the assistant makes afterward.

ConservativeAggressiveLong-term planningFixture chasingUnbiasedFavor own clubStrong Starting XIBalanced 15Fixed formationFlexible formation
Fig. 1: A one-time onboarding screen. Each preference sits on a spectrum, not a checkbox.

02

Weekly Assistant Manager report

Before every gameweek, users receive a personalized report summarizing what went well and wrong the previous week, plus recommended transfers, captain and vice-captain, starting XI and bench order, chip strategy, and projected points.

Every recommendation follows the same structure: Recommendation → Justification → Confidence rating. Rather than just naming a player, the Assistant Manager explains why, referencing fixture difficulty, form, injuries, expected minutes, and the user's stated preferences.

HomePick TeamTransfersLeaguesFixturesThe ScoutTransfersGameweek 14 · Deadline: Sat 09:00£4.5mGK£6.0mDEF£4.5mDEF£4.0mDEF£5.0mDEF£10.0mMID£5.5mMID£10.5mMID£6.5mMID£6.5mFWD£8.0mFWDAssistant ManagerGW14 ReportGW14 report · tuned to your styleHi Vir, here's what I'd focus on forGameweek 14. Given your aggressive,fixture-chasing style, I'd captainHaaland at home to Burnley: his xG is0.82 and City are heavy favourites,so it fits how you like to swing forreturns. On transfers, Bowen has atough three-week run, and Palmer'sfixtures soften from GW15 onward, somoving Bowen → Palmer this week linesup nicely with your 3-GW horizon.For the bench, I'd float Konsa aboveEstève since Estève has real rotationrisk after Wednesday's cup tie.I'd hold your Wildcard for now,DGW17 is shaping up to be a muchstronger week to play it.Make transferEdit benchAsk a follow-up…↑
Fig. 2: The weekly report is docked into the right-hand panel of the Team Management page, right where Player Selection normally sits. A dashed cue points from the pitch to the captain recommendation.

03

Conversational assistant

Alongside the weekly report, managers can ask follow-up questions through a natural-language chat interface, "Why captain Haaland over Saka?", "Should I take a four-point hit?", "Why are you saving my Wildcard?" Because the assistant remembers each user's philosophy, responses stay personalized rather than simply optimizing projected points.

HomePick TeamTransfersLeaguesFixturesThe ScoutTransfersGameweek 14 · Deadline: Sat 09:00£4.5mGK£6.0mDEF£4.5mDEF£4.0mDEF£5.0mDEF£10.0mMID£5.5mMID£10.5mMID£6.5mMID£6.5mFWD£8.0mFWDAssistant ManagerAsk a follow-upChat · threadWhy captain Haaland over Saka?Assistant ManagerBecause your setup rewards ceiling oversafety. Haaland is home to Burnley withan xG of 0.82 and City are heavyfavourites, so his projection is 8.4.Saka has a tougher away trip at a top-4side and projects 5.9, so on youraggressive style Haaland is the higher-upside pick this week.Take a -4 hit for Palmer?On this gameweek alone the -4 ismarginal, but over your usual 3-weekhorizon Palmer's fixtures soften a lot,so taking the hit still nets around+6.1 points versus keeping Bowen.Given your fixture-chasing lean, I'dtake the hit.Ask the Assistant Manager…↑
Fig. 3: Tap "Ask a follow-up" on the report and the same panel switches into a chat thread. The pitch stays put, so the user can react to a suggestion without leaving the page.

04

Explainability & trust

Football is inherently unpredictable, so the Assistant Manager presents every recommendation with an associated confidence rating instead of implying certainty. This communicates uncertainty transparently while helping users understand which recommendations are strongly supported by the available data and which involve greater risk.

Monetization

Assistant Manager is offered as an optional premium feature when users create their team at the start of each season.

Managers who subscribe gain access for the entire season, creating an annual subscription model that aligns naturally with FPL's seasonal gameplay. Teams using Assistant Manager display a small badge beside their squad, providing transparency while preserving competitive fairness.

Projected solution impact

Based on existing user behavior and market trends, the proposed feature has the potential to deliver value for both managers and Fantasy Premier League.

Less research

Reduces the research burden for casual managers while keeping experienced players in complete control.

Personalized

Creates a personalized experience within an otherwise standardized game.

Season-long engagement

Encourages greater season-long engagement through weekly AI reports and ongoing conversations.

Recurring revenue

Introduces a recurring seasonal revenue stream without restricting core gameplay, and lays a scalable foundation for future AI-powered features across FPL.

Reflections

What went well

The project successfully identified a meaningful opportunity to apply AI without fundamentally changing the existing product. By focusing on personalization rather than automation, the Assistant Manager complements the traditional FPL experience instead of replacing it.

Potential next steps

While this concept focuses on delivering weekly recommendations, future iterations could make the Assistant Manager even more adaptive.

  • Learn from user responses to previous recommendations, refining future advice based on accepted or rejected suggestions.
  • Incorporate live team news and late injury updates before each gameweek deadline.
  • Grow increasingly personalized as it learns each manager's decision-making habits throughout the season.

Each of these stays tightly aligned with the core idea: an AI that becomes a better assistant by learning how you like to manage your team.

Key takeaway

One of the biggest lessons from this project was that AI is most valuable when it helps users make better decisions, not when it makes decisions for them. Rather than optimizing purely for projected points, the Assistant Manager adapts to each user's individual management philosophy, creating a more personal and engaging Fantasy Premier League experience.

Up next

Avataar, mobile 3D capture  

Redesigning a mobile 3D capture workflow from six minutes to under 150 seconds.