Block Management Software: Service Charge Queries and Repairs With AI

Block management software should hold the lease and unit records, run service charge budgets and accounts, track repairs from report to invoice, and give leaseholders a portal. AI can answer routine status and balance queries and triage repairs from photos, but formal notices, demands and disputes should stay with a person.

Which AI you get depends mostly on the platform you already run, so check what it offers before buying anything. Nothing here is legal advice: the rules for notices, consultation and recovery differ by place and lease, so have a qualified adviser confirm anything statutory. Every figure in the examples is illustrative.

Follow me on Instagram@sagnikteaches

Six jobs the software must do before AI matters

The pattern in the last two columns repeats down the table. AI reads, sorts and drafts, and a person decides and signs. A simple test decides which side a task falls on: if an error could lead to a tribunal, an unrecoverable cost or a right the leaseholder can exercise against the client, it stays with someone who understands the lease. I use that test throughout.

Connect on LinkedInSagnik Bhattacharya

Sorting leaseholder queries into three routes

Start by sorting your own inbox, not by shopping. Take 200 queries from the last quarter and classify each by what the answer depends on. Here is an invented sample for a managing agent with 40 blocks, with the route each type would take.

Subscribe on YouTube@sagnikteaches

That is 37 per cent on route A, 48 per cent on route B and 15 per cent on route C. The proportions will differ for you. What does not differ is the shape: a third or so is plain fact retrieval, about half needs a draft with a human in the loop, and a small slice is not for software. Count your own before you believe anyone's brochure.

A general assistant can do the first sort for you, on anonymised text. Here is a prompt and an illustrative reply.

Classify each leaseholder message below into one type: status of a known job, document request, balance query, new repair report, charge explanation, lease question, complaint or other. Then give the route: A (auto-answer from system data), B (AI drafts, person sends) or C (person only). If unsure, choose C and say why.

1. "Has anyone booked the roofer yet? The stain is getting bigger."
2. "Can I paint my front door a different colour?"
3. "I got a letter saying I owe more than last year. Why?"

It put message 1 on route B rather than A: a growing stain is new information about an existing ticket, so a person should look at the photo. It labelled message 2 a lease question on route C, correctly. For message 3 it chose B and added "check the budget apportionment before replying", which is the sensible caution. The routing was right on all three, and the one place I would override it is the first, because it noticed the escalation that a status template would have missed.

Repair triage: from a vague message to a usable ticket

Repair triage is where block management and AI fit best, because the input is messy and the output is a structured ticket. Fixflo's Aidenn assistant, for example, is described on Fixflo's own page as analysing photos, diagnosing faults and escalating, in a large number of languages, and as producing a job report after each conversation with onward works and links to parts. It lets the agent upload repair policies, asset information and emergency contacts. That page also describes a separate Aidenn Compliance tool for reading fire risk assessments. These are vendor claims, so ask to see a triage on your own photos.

The aim is the same whatever you use. Here is a before and after for the same report.

Notice what the ticket does not say. It does not decide who pays, because that depends on the cause and the lease. A tool that confidently assigns responsibility has gone past what it can know. I would also write the emergency rules down for the tool: for instance, anything involving gas, active water near electrics, a person trapped in a lift or loss of fire protection goes straight to a phone call, whatever the AI's confidence. The triage patterns in how property managers use AI to triage maintenance requests include a priority ladder you can adapt.

Service charge accounts: let AI read, never approve

The accounting side offers the least glamorous and most dependable AI gain: reading invoices. A supplier-written directory entry for Inox describes AI invoice processing that reads, categorises and prepares bulk-uploaded invoices for approval, along with automated bank reconciliation and anomaly spotting. As with every vendor claim here, ask for a demo on a messy real invoice.

A quick sum shows why it matters. Say an agent processes 600 invoices a quarter across 40 blocks. Coding one by hand takes 3 minutes. Checking an invoice that the AI has already read and coded takes 1 minute. That saves 2 minutes on each, or 1,200 minutes, which is 20 hours a quarter, about $700 at $35 an hour. The AI never approves a payment. A person confirms the supplier, the block it belongs to and whether the work was ordered.

Where AI should not go is anything that issues a demand or starts a process with a legal clock. Late or incorrectly worded notices can leave costs unrecoverable, and the consequence lands on the agent and the client, not the software. Use AI to prepare for those jobs, for example to scan a year of correspondence for the likely questions before a budget goes out. Keep the notice templates current and the sign-off human.

Four products compared on what their own pages say

This table sticks to what I could read on vendor or supplier pages in November 2026. It is a starting list for demos, not a ranking.

Two points come out of the table. First, each AI claim covers one step: Fixflo's is aimed at repairs, Inox's at invoices and MRI's at reading leases, so the "AI" label tells you very little until you see the feature. Second, none of the pages I read promises AI that answers leaseholder messages by itself. If that is your goal, you are either adding a help-desk agent on top, or drafting replies with a general assistant. The mix-and-match approach is covered in handling tenant queries without extra staff and AI email triage for professional firms.

Costing an AI query agent for 900 queries a month

Take a managing agent with 40 blocks, 1,800 units and 900 leaseholder queries a month, with each query costing 6 minutes of staff time at $35 an hour. Use the routes from the sample above. Route A is 37 per cent, so 333 queries. Route B is 48 per cent, 432 queries. Route C is 15 per cent, 135 queries.

On cost, a help-desk agent priced per outcome, such as Intercom's Fin at $0.99 an outcome, would charge $329.67 for 333 route-A queries. Intercom also counts completed workflows, hand-offs included, as outcomes, so any route-B message Fin passes to a person can add to that figure. Add four business-plan assistant seats at $25 each for the drafting on route B and the monthly total is about $430, leaving roughly $1,490 of net value a month. These are illustrations built on invented volumes, so rerun them with your inbox.

One trap sits inside that price. The vendor counts an "assumed resolution" when the customer goes quiet for 24 hours after the AI's last answer. A leaseholder who gave up has been counted as helped. If 20 per cent of the 333 route-A queries were really silent failures, that is 67 queries that will return as repeat contacts or complaints. The check is to track repeat contact on the same issue within seven days, and to sample 20 closed conversations a week against the ticket record.

An answer policy the AI works from

Whatever tool answers, give it written rules. A short policy is easier to audit than a long prompt. Here is a filled-in version to adapt.

Reply rules for [firm name]

MAY: quote job status, dates and contractor names exactly as in the ticket; send stored documents; explain a budget line using the published budget notes; ask for photos and details.
MUST: confirm the unit and the email on file before discussing a balance; say "I don't know" and hand over when unsure; pass every complaint, lease question, dispute and formal notice to a named person.
MUST NOT: promise a date that is not in the ticket; say who is responsible for a repair; interpret the lease; give legal or financial advice; discuss another leaseholder's account; call a charge valid, capped or recoverable.

Here is the kind of slip the policy catches. A draft reply read: "The lift contractor has been booked for Thursday and the lift will be back in service by the end of the week." The ticket showed the Thursday booking but no return-to-service date, so the second half was invented. The fix is the first MAY rule: quote the ticket, nothing else. Edited, the sentence says the contractor is booked for Thursday and that you will update the leaseholder when the repair is confirmed.

A reply that went past its remit

Picture an agent whose assistant told a leaseholder, "your increase is within the permitted limit." The statement looked helpful and nobody had checked it. When the leaseholder later challenged a demand, the agent had a written reply to explain. It showed up as a complaint, not as a software error. That is why the policy bans words like valid, capped and recoverable, and why the person reading route-B drafts is looking for exactly those words.

Explaining a budget line without opening a dispute

"Why has my charge gone up?" is the query most likely to turn into a complaint, so it belongs on route B. The draft is only as good as the figures you give it. Paste the budget note, not the lease, and tell the assistant what it may not say.

Explain this budget line to a leaseholder in plain English, under 100 words. Do not say whether the charge is fair, capped, valid or recoverable, and do not mention any other leaseholder. If the note does not explain the rise, say a colleague will follow up.

Budget note: Lift maintenance contract, annual cost $9,600 (last year $8,400); renewed after retender; 48 flats; apportioned by the lease schedule.

An illustrative reply explained that the lift contract had been retendered and renewed at $9,600 a year against $8,400 the year before, a rise of $1,200, shared between the 48 flats under the lease schedule, and offered a colleague's follow-up. It also added "this is a reasonable increase given market conditions", which nobody had supplied and which a challenger could quote. I would delete that clause. The $1,200 is a fact from the note. The judgement is not the AI's to make. If every flat paid an equal share, the rise would be $25 a flat, but the lease schedule may split it differently, so the draft rightly avoided quoting a per-flat figure.

Out-of-hours reports and emergencies

Overnight and weekend reports are where AI can help most and fail worst. Help, because someone is awake at 2 a.m. to ask for photos and note the time. Fail, because a worrying leak read as a routine job costs hours. Give the tool a short list of phrases and situations that always trigger a phone call to the on-call person, with no triage conversation first: a smell of gas, water near electrics, a fire alarm or smoke, a person stuck in a lift, a broken entry door that leaves a block open, and any report that mentions a vulnerable resident.

Then test it. Send ten invented overnight messages, two of which are true emergencies phrased vaguely ("it's all wet in the hall and the lights are flickering"). Count how many reached the on-call person within five minutes. If the answer is not ten out of ten, tighten the rules before it goes live. The priority ladder in the maintenance triage tutorial linked above is a good starting point. Record who was called and when, so that a later complaint can be answered from the ticket.

A demo script for any block management vendor

Vendors demo well on tidy data. Ask for these seven things on your own examples, and score each as shown, partly shown or not shown.

1. A leaseholder reports a leak with two photos. Show the triage, the priority it chose and the ticket it created.
2. Show a leaseholder asking for a job status and the reply, with the source of every fact visible.
3. Show the audit trail: what the AI was given, what it produced and who approved it.
4. Show an invoice read by AI with a deliberately wrong block, and how a person corrects it.
5. Show the portal as a leaseholder sees it, including how they reach a person.
6. Show how a lease question is routed, and prove that the AI did not answer it.
7. Show the export: all tickets, accounts and messages, in a usable format.

Score them on the demo, not on the brochure. Where a vendor says a feature is "coming soon", price the product without it. Add the quoted software fee, any per-unit charges, onboarding and the cost of loading your existing records, and compare quotes on the same list of lines.

Rolling it out in four steps

1. Week 1: classify 200 past queries, as in the table above. Count your own routes.
2. Week 2: turn on route A for one query type, such as job status, in one block or for one team. Keep a person reading every reply.
3. Week 3: add document requests, and start drafting route B replies for the team to edit. Track edits: if people rewrite more than a third of drafts, fix the template, not the people.
4. Week 4: audit 20 replies from each route against the system record and measure repeat contact within seven days. Decide what to widen and what to pause.

Start with the platform you have. Check whether the AI already in your system covers a step before you add another subscription, and for out-of-hours calls weigh an AI line against a human answering service before you pay for either. Then widen only where the audit says the replies were right. The measure that matters is not how many queries the AI closed. It is how many leaseholders did not need to ask again.

Block management and AI: questions managing agents ask

Can AI answer leaseholder queries about service charges?

It can answer routine ones from live data, such as a balance, a payment date or the status of a job. It should not interpret a lease, defend a demand or give legal advice. Draft anything that explains why a charge rose and have a person approve it before it is sent.

Do block management platforms include AI for repairs?

Some do. Fixflo's Aidenn assistant analyses photos, diagnoses faults and escalates, and a supplier-written Inox listing describes AI invoice processing. Others, such as the Blocks Online portal pages I read, mention a problem-reporting button and ticketing but no AI. Ask each vendor for a demo on your own sample queries.

What should never be automated in a managing agent's inbox?

Formal notices, service charge demands, consultation steps on major works and anything that could become a dispute. A good test: if an error would cost money that cannot be recovered or give a leaseholder a right to act against the client, a person who understands the lease writes and sends it.

How do I measure whether AI replies are safe?

Audit a sample every week. Pull 20 automated or drafted replies, check each against the system record, and log any wrong fact. Pause that query type if more than two are wrong. Also track repeat contact within seven days, which shows answers that looked complete but were not.

Further reads

• How Much of a Property Manager's Week Can AI Take Over? — How much of a property manager's week AI can realistically take.
• AI Landlord Updates and Owner Statements in Minutes — Owner and landlord updates drafted from the same records.
• Can AI Triage Emergency Plumbing Calls Out of Hours? — Emergency triage rules from a trade that lives with them.
• Questions to Ask Before Buying AI That Touches Client Data — Data questions to put to any block management vendor.
• Use the AI Already in Your Practice Software Before Buying More — Use the AI in your current system before adding another.
• Your First 30 Days of AI Customer Support for an Online Shop — A 30-day support rollout pattern you can borrow.
• AI Tools for Commercial Property Managers: Service Charges, Leases — Three jobs for a commercial property manager: service-charge replies from a building pack, a verified lease-event sheet and owner-report drafts.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Fixflo's Aidenn, Aidenn Compliance and block management pricing pages, MRI Software's block management page, Blocks Online's property portal and ticketing pages (read 1 November 2026), a supplier-written Inox directory entry, and the series fact base for Intercom Fin pricing.