01 / THE EXAMPLE
The hall is closed on Wednesday.
Who finds the replacement?
Consider an invented example of an ordinary problem. The community hall where a club trains is closed for one evening. The club needs a one-hour replacement session. The hall can offer Wednesday at 18:00 or 20:00, both for €60. Most of the group prefer 20:00, after work. One athlete has to leave by 19:00 to catch the last bus home.
Picture how this is handled today. A coach or volunteer coordinator reads the message from the hall, checks the options and asks the group. If they know about the last bus, they can ask whether 18:00 works for the group or discuss an agreed transport alternative. People can miss this information too. The point is to make the session’s limits visible to whoever handles the change.
A real allocation process. NFF Oslo’s published futsal guidance describes clubs submitting hall preferences, coordination between sports and a final allocation meeting. It also reserves weekday training before 20:00 for children and young people. [6] Our example is not an Oslo booking. It illustrates a choice that would first have to fit the local allocation rules.
The club’s AI assistant handles it instead. Its instructions: find a one-hour slot, at most €60, preferably 20:00. It books 20:00. Every instruction was followed. The athlete cannot come.
That is the practical question this report explores through the ordinary arrangements that keep sport running: training slots, pitch swaps, transport and equipment. When software starts making those arrangements, the result depends on two things: what it was told, and how much it is allowed to do on its own.
A note on words: by AI assistant (often called an AI agent) we mean software that can use an AI model, information and tools such as email or a booking system to carry out a task for someone. When we say assistants “negotiate”, we mean they exchange messages and offers. They act on assigned instructions; that does not guarantee a good result.
02 / WHAT IS ACTUALLY CHANGING
From answering questions
to making arrangements.
In a chat window, you ask, it answers, and you decide what to do. The change now under way is that assistants are being connected to tools and to each other, so that they can carry out tasks rather than only give advice. Three developments matter for sport.
TECHNICAL STANDARDAssistants from different companies can deal with each other.
The Agent2Agent (A2A) protocol is an open standard that lets assistants built by different vendors find out what each other can do and exchange tasks and messages. [1]
Our reading for sport: assistants from different suppliers could communicate if both implement the standard and have the necessary access. Connecting an existing municipal booking system would still take integration work and permission. People would still set the rules for who gets a scarce slot.
RESEARCH · ANTHROPIC · SEPTEMBER 2026They can close a deal and still miss what the person wanted.
In Anthropic’s Project Swap, AI assistants traded books on behalf of employees after a short chat about what each person liked to read. The researchers traced most of the shortfall in results to the assistants’ imprecise picture of each person’s preferences. It was a controlled test, run with the company’s own models and staff. [2]
Our reading for sport: “find us a session” hides limits of travel, money, work and family. A booking that fits the request but not the person is not a success.
RESEARCH · ANTHROPIC · AUGUST 2026Several assistants together do not automatically work well.
In Anthropic’s experiments with groups of assistants, some groups coordinated poorly, some assistants simply went along with the others, and assistants with conflicting goals clashed. Behaviour differed between models. [3]
Our reading for sport: testing one helpful assistant does not establish what happens when the club’s, the hall’s and another club’s assistants compete for the same Wednesday evening.
The research raises questions worth testing in sport. It does not establish how a real club or hall would perform. The evidence limits are collected in the method note at the end.
03 / TRY IT
How much of the decision
would you hand over?
The same Wednesday, three levels of authority. First choose how much the club lets its assistant do. Then decide whether the athlete’s travel limit is part of its instructions.
These are three steps in our illustration, not a standard classification. In a real trial, the club and facility would need to agree when an offer becomes a booking and what can be changed. A club’s instructions cannot replace the facility’s allocation rules.
Suggest
Finds options. A person decides and books.
“Here are two possible slots.”
Negotiate
Agrees a provisional offer. A person approves.
“The hall can hold 20:00 for us.”
Confirm
Books within limits the club set in advance.
“The booking is confirmed.”
EXAMPLE · Invented times and prices, scripted results, no live AI
ONE HALL · THREE INTERESTS
Keep training goingThe hall
Fill a free slotThe athlete
Get home afterwards
SUGGEST · NOTHING IS SENT
A sensible answer. Incomplete instructions.
- Club’s instructionsFind a one-hour slot, at most €60. Prefer 20:00.
- Assistant’s suggestionWednesday, 20:00 to 21:00, €60.
- Next stepA person checks the suggestion before contacting the hall.
Nothing is booked.
The suggestion matches the instructions, but the athlete would miss the last bus. The person checking may catch this, if they know about it.
An instruction can be followed perfectly
and still leave someone behind.
Two lessons follow. First, the instructions matter as much as the authority: a €60 limit protects the club’s budget but says nothing about whether everyone can attend. Second, a human check only helps if the person checking has the right information and enough time. Before giving an assistant more authority, a club should be able to answer three questions: what must it always know, what may it decide alone, and who can overturn its decision?
04 / LOOKING TO 2030
Three ways
this could develop.
The same technology could lead to very different local arrangements. These three sketches are for discussion, not forecasts. A local approach could combine elements of more than one sketch.
SKETCH 1 / A SHARED LOCAL SYSTEM
The municipality and clubs
agree common booking rules.
Clubs, halls and the municipality agree how slots are requested, swapped and confirmed. A club can use its own assistant, a shared one provided locally, or simply a person. Every booking shows who authorised it. The municipality funds support for clubs that cannot run a system themselves.
What gets easierLess repeated admin, easier swaps between clubs, and support for small clubs that they could not afford alone.
What stays hardSomeone has to pay for and maintain the shared system, and agree whose needs come first.
Signs to look for: public allocation rules, a funded option for clubs without an assistant, and booking records that clubs can take with them.
SKETCH 2 / A PAY-FOR-PRIORITY PLATFORM
One commercial platform runs
bookings, transport and equipment.
Its assistants make arrangements quickly. Priority access and stronger negotiating features cost extra. Small clubs like the convenience but cannot see why another club got the better slot, or what they would lose by leaving. Every club may have an assistant, yet the clubs that pay more still get the best times.
What gets easierFast, convenient booking through one familiar app.
What stays hardAllocation rules are hidden, decisions are hard to challenge, and switching provider means losing history.
Signs to look for: paid priority for public facilities, closed booking systems, and clubs unable to see why a request was refused.
SKETCH 3 / ASSISTANTS PREPARE, PEOPLE DECIDE
Assistants search and draft.
People confirm.
Clubs and municipalities let assistants find options and prepare arrangements, but a person confirms anything that affects who can take part or what it costs. Assistants are given more authority one task at a time, once they have shown that they help.
What gets easierClubs learn what assistance is good for while decisions stay visible, and human checks can catch needs the assistant never saw.
What stays hardChecking takes time. Clubs with few volunteers may face delays, and approvals can become a box-ticking routine.
Signs to look for: short written limits on what assistants may do, time set aside for checking, and clear reasons for keeping decisions with people.
A RISK IN EVERY SKETCHWhat happens if a shared booking service fails on a busy weekend?
If many clubs rely on the same service, one outage or cut connection could disrupt arrangements across a whole area. Clubs and municipalities should know in advance how to find the last confirmed booking and carry on by phone, email or a shared calendar. This is a planning question, not a report of a real incident.
05 / WHAT EACH ROLE CAN DO NOW
Decisions to make
before the software arrives.
Start with agreements between people. They need time, but do not require a software purchase. These are our suggestions for each role.
Write down what you know.
- Ask participants which practical limits a session must meet, such as finishing by 19:00, step-free access or no extra fee.
- Agree what the booking coordinator needs to know. Share the session requirement rather than a named person’s financial, medical or family circumstances; keep sensitive discussions with the appropriate person.
- Ask to see any booking change that affects who can attend before it is final.
Set the club’s limits.
- Agree which tasks an assistant may only suggest, which it may negotiate and which it may confirm.
- List the limits it must respect: budget, times, travel and accessibility.
- Name a person and a backup who can ask the facility to correct or cancel a booking. Check deadlines and charges before relying on cancellation.
- Keep one agreed booking record showing what was booked, who authorised it, and whether it is provisional or confirmed.
Keep allocation rules public.
- Publish how slots in public facilities are allocated and how a club can challenge a decision.
- Make sure clubs without an assistant can still book on equal terms.
- Ask suppliers whether priority can be bought, and whether clubs keep their booking history if they switch.
- Plan how bookings continue if a system fails.
Help clubs find common ground.
- Start with shared guidance and a named contact for clubs. Consider a shared service only if there is capacity to fund and maintain it.
- Prepare template wording on what an assistant may do for a club.
- Collect real examples from clubs before endorsing any tool.
In each case, the aim is to settle these questions before a supplier’s default settings quietly become local practice.
06 / HOW TO TELL IF IT HELPS
Faster is not
the same as better.
Sport England’s account of volunteer experience highlights requests for reasonable, flexible time demands, less responsibility and additional support. [7] This is English context, not a measure of every European club. Our practical test is whether a tool reduces total work rather than moves it to another volunteer.
If a club or municipality tries an assistant, speed alone will not show whether it helped. We suggest judging it on three questions.
Whose work actually shrank?
Count the time spent checking, correcting and explaining, not only the time saved preparing.
Who can still take part?
Check travel, cost and accessibility, and whether people without an assistant still have a fair route.
Can someone object?
People should be able to see why a decision was made and know who can change it.
Then compare who could use the arrangements before and after. Did the time saved make room for more people to take part? Did anyone face extra cost or work?
This matters because sport already reaches some people more than others. The EU Work Plan for Sport 2024 to 2027 links digital innovation with participation, volunteering, good governance and accessible facilities. [4] The European Commission’s September 2026 summary of the latest sport Eurobarometer reports that younger people, people with more education and people with fewer financial difficulties take part more. [5] Our view is that any time saved should be judged alongside who can take part and who still has to do the work.
A club may also find that a shared calendar, clearer rules or a phone call solves the problem with less effort. That is a useful result too.
07 / A 90-DAY TRIAL
Start with one hall
and one awkward decision.
A club can start on Monday with one recent booking change: write down who handled it, the time spent and whose needs emerged late. The 90-day outline below is an optional way to take that further with a facility and other clubs. It is our proposed exercise, not a tested programme.
Look at what really happens.
Name a club coordinator and a facility contact, agree how much time they can give, and go through recent booking changes with the people involved. Where did the work pile up? Which needs came to light too late? How were disagreements settled? Include at least one person who usually has little say.
Result: one well-documented problem and a starting point to measure against.
Run it on paper.
Use invented versions of those situations in a tabletop session. Compare the current process, a simple shared calendar and a worked example of an assistant’s proposal. Use fictional people and constraints; no live AI is needed. Change the assistant’s authority, leave out one person’s need, and ask participants to challenge an answer that looks good.
Result: a tested list of what an assistant must know and may decide.
Decide whether a real trial is worth it.
Compare total effort, missed needs and how easily mistakes could be undone. If a technical trial makes sense, first agree its cost, staff or volunteer time, permitted system access and handling of personal information with the facility. Start with suggestions that a person checks; consider more authority only after reviewing the results.
Result: a brief for a small trial, or a clear reason to choose a simpler solution.
Before any live trial, agree what would count as an improvement and what would stop it: for example, the same person’s need being missed more than once, more checking work than before, or no clear way to undo a booking.
THE QUESTION TO TAKE FORWARD
How can AI assistants help sport organisations make arrangements with each other, while the people affected keep a real say in the decisions?
Before AI assistants start booking the hall,
who will set the rules?
08 / ABOUT THIS REPORT
What this report is,
and what it is not.
How it was made. This desk-based report from Sport Singularity uses a technical standard, two company research publications, official EU sport sources, NFF Oslo’s allocation guidance and Sport England’s account of volunteer experience. Sources were checked on 11 November 2026.
What is ours. The booking example, authority levels, 2030 sketches, role suggestions and 90-day exercise are our illustrations and proposals. The interactive example has invented times, prices and scripted results; no AI runs in it. The provider research does not establish effectiveness in sport. The Oslo and English examples are local context, not evidence of conditions across Europe.
What remains to be tested. These sources do not establish that assistants improve bookings in European grassroots sport or that the arrangements in our sketches are being used there. No club interviews, pilots or independent AI tests were carried out for this report. A local exercise should establish whether the problem is important enough to address and whether a simpler approach works better.
AI assistance. AI tools assisted research, drafting, translation and structured editorial checks. The two AI experiments are reported by the same model provider.
- TECHNICAL STANDARD · CHECKED 11 NOV 2026
A2A: What is Agent2Agent?
Used for how assistants from different vendors discover each other’s capabilities and exchange tasks and messages. Describes a standard; says nothing about use in sport.
Used in: What is actually changing - COMPANY RESEARCH · 24 SEP 2026 · CHECKED 11 NOV 2026
Anthropic: Project Swap
What happens when agents trade for us? Used for the finding that most of the shortfall came from an imprecise picture of people’s preferences. A controlled test by the company that builds the models; not tested in sport.
Used in: What is actually changing - COMPANY RESEARCH · 13 AUG 2026 · CHECKED 11 NOV 2026
Anthropic: Patterns and problems in emerging multiagent systems
Used for poor coordination, conformity and conflicting goals in groups of assistants, and for differences between models. Experimental settings limit how far the results apply.
Used in: What is actually changing - OFFICIAL EU SPORT POLICY · 2024 TO 2027 · CHECKED 11 NOV 2026
Council: European Union Work Plan for Sport
Resolution C/2024/3527, paragraph 15 and Annex I. Used for the EU’s priorities on participation, volunteering, governance, digitalisation and accessible facilities. Policy priorities do not show demand for AI assistants.
Used in: How to tell if it helps - OFFICIAL SURVEY SUMMARY · 22 SEP 2026 · CHECKED 11 NOV 2026
European Commission: Sport and physical activity Eurobarometer
Commission summary of the latest Eurobarometer on sport, used for differences in participation by age, education and financial situation. A participation survey, not a study of AI.
Used in: How to tell if it helps - SPORT FEDERATION GUIDANCE · PUBLISHED 8 AUG 2024 · CHECKED 11 NOV 2026
NFF Oslo: Allocation of futsal training and event time
Published local process: hall preferences, coordination between sports and an allocation meeting; weekday time before 20:00 reserved for children and young people. This is Oslo futsal guidance, not a Europe-wide rule or an AI deployment.
Used in: The example - SPORT INSTITUTION · CHECKED 11 NOV 2026
Sport England: Taking an honest look at the volunteer experience
An institutional account by its head of volunteering, used for volunteers’ requests for manageable time demands, less responsibility and support. English context, not an AI study or a new European prevalence estimate.
Used in: How to tell if it helps