An inbox that routes itself
Prismlet reads each email or ticket as it lands and answers who should own it, what the sender wants and how urgent it is. Your own system uses the answer to send it to the right queue, and a person checks anything the answers are unsure about.
Ask for an inviteToday
- Every email, ticket and web form lands in one shared box, and someone reads each one before it goes anywhere.
- Keyword rules catch the word, not the request. For example, a firm order to bind cover gets filed under whatever subject line it landed on.
- Cargo held by customs, a bind order, or a concern that someone may be at risk of harm sits in the pile in arrival order, next to a newsletter.
- A second matter folded into the same message, a missing account number, or a request to speak to a person is easy to miss on a quick read.
- A shared inbox that falls behind, for example during a storm or a busy week, can take days to clear.
What changes
- Every email or ticket gets who should own it, what the sender wants and how urgent it is, as it lands, and each answer says how sure it is.
- Your system can put the urgent ones first, and a person checks anything the answers are unsure about.
- A bind order, cargo held by customs, or a concern that someone may be at risk of harm is flagged as it lands, for a person to act on straight away.
- A second matter in the same message, a missing detail, or a request to speak to a person is flagged too, so it is not lost inside a longer email.
Try it on this example
Subject: Charged twice + app broken
Ticket text, or the whole thread
- Which team should handle this support ticket?Billing and subscriptions100%
- What does the customer mainly want in this support ticket?Report a problem100%
- How urgent is this support ticket, from what the customer says?High100%
- Does the customer ask to speak to a person, a manager or someone by phone?No90%
- Does the team need details from the customer that the ticket does not give, before it can start?No74%
- Does the ticket raise a second issue that belongs to a different team?Yes83%
- Does the customer describe a device or charger getting hot, smoking, swelling, sparking or hurting someone?No91%
These are real answers stored from one run on this example.
Email subject: RE: Quote TRU/PO/88213 Tolworth Street Properties Ltd
Attachment file names: EICR_14_Tolworth_Street_Aug2026.pdf; Signed_Statement_of_Fact_Tolworth_Street.pdf
Email body
- What is the broker mainly asking us to do in this email?Bind order or hold covered99%
- Does the broker ask us to bind, go on cover or hold the client covered, anywhere in the email?Yes99%
- Does the broker say cover or proof of cover is needed today, or that the client is uninsured until we act?Yes96%
- Does the broker chase something they asked for or sent before?Yes98%
- Does the email supply information we asked for as a condition of a quote or of cover?Yes98%
- Does the email report a claim, a loss or a circumstance that may lead to a claim?No91%
- Does the broker or the client express dissatisfaction with our service, terms or decisions?No87%
- How soon must someone act on this broker email?Immediate95%
- Is information missing that the desk needs before it can act?No62%
- Does the broker ask for a phone call or to speak to a person?No91%
These are real answers stored from one run on this example.
Subject: RE: Booking HAM-58213 / 1x40HC Ningbo - Hamburg - URGENT
Attachment file names (empty if none):
Latest message, without the quoted thread
- What does the sender mainly want from us in this email?Status chase100%
- Which transport mode does the email concern?Ocean, full container100%
- How soon must someone act on this email, from what the sender says?Critical100%
- Does the email refer to a shipment already booked with us?Yes98%
- Does the email say cargo is held, selected for inspection or stopped by customs or another authority?Yes97%
- Does the email say a cut-off, sailing, flight or delivery slot has been missed or may be missed?Yes98%
- Does the email mention dangerous or hazardous goods?No93%
- Does the sender express dissatisfaction with our service?Yes98%
- Does the email raise a second matter that a different desk would act on?Yes96%
- Is information missing that the desk needs before it can act?No78%
- Does the sender ask for a phone call or to speak to a person?Yes99%
These are real answers stored from one run on this example.
Test it on your own data
The quickest way to decide is a file run on your own records. A file run in the app answers up to 10,000 rows and gives you the file back with the answers added.
This is for the first prism, Sort a shared inbox by team, urgent ones first. Each prism lists its own fields in "The prisms behind it" below.
- What to send
- A CSV or TSV file with column names in the first row, or a JSON array of objects, one item per row. You pick the column for each thing the prism reads: Subject; Email, form or ticket text. A row's fields together take up to about 100,000 characters.
- What comes back
- The same rows with the prism's columns added: service, service_probability, request_type, request_type_probability, urgency, urgency_average, person_at_risk, person_at_risk_probability, homeless_or_threatened, homeless_or_threatened_probability, complaint, complaint_probability, member_enquiry, member_enquiry_probability, second_matter, second_matter_probability, missing_details, missing_details_probability, callback_requested, callback_requested_probability. Each answer says how sure it is, so the unsure rows are easy to find and give to a person.
- Against your team's answers
- Add columns with the answers your team gave, and the app reports how often Prismlet agreed, question by question and at each cutoff, split by a column such as language or site, with every miss listed. Your team's answers stay in your browser.
How it reaches your team
Works today
- Call the API from your own system as each email or ticket arrives
- Run a file of up to 10,000 emails or tickets in the app, one item per row, up to about 100,000 characters a row
Built with you
- We work with you to send each email or ticket from your inbox or helpdesk, such as Outlook or Zendesk, as it arrives
- We work with you to put the answer back where your team works, as a category and folder, a queue, or a column in your tracker
- We work with you to alert the right desk straight away when something needs same-day attention, such as a message to the duty coordinator
The prisms behind it
Sort a shared inbox by team, urgent ones first
Fields
- Subject
- Email, form or ticket text
Context
We are Westmere Council, a UK unitary council. Residents, businesses, councillors and other organisations write to one contact address, through the website "contact us" form, or into the customer services case system. Each message is read here on arrival and sent to one service's queue. A person in the customer services triage team handles anything the answers are unsure about. Routing rules: - One message, one service. When a message raises more than one matter, route it by the matter that would do the most harm if it waited. A concern that someone may be at risk of harm always comes first. - A concern about an adult or a child is passed to a trained person in adult social care or children's services, who decides what to do. Nothing here decides that anyone is at risk. - Councillors and MPs who write for a resident are members, and their enquiries have their own deadline. - Requests for recorded information (Freedom of Information, Environmental Information Regulations) or for someone's own personal data go to information governance, whichever service they are about. - The case system matches references, addresses and account numbers. Nothing here looks them up.
Questions
Which council service should handle this email or ticket? Choice
Choose the one service that should handle the message, using the routing rules in the context. When the message raises more than one matter, choose the service for the matter that would do the most harm if it waited; a concern that someone may be at risk of harm always outranks the others.
What does the sender want from the council in this email or ticket? Choice
Judge the matter that would do the most harm if it waited, as in the routing rules in the context. Dissatisfaction elsewhere in the message is recorded by the complaint question.
How urgent is this email or ticket, from what the sender says? Scale
Rate the most urgent matter in the message, whatever service it belongs to. Do not compare dates with today; rely on words such as "tonight", "tomorrow" or "since last week".
Does the message raise a concern that a child or an adult may be at risk of harm or neglect? Yes / No
Count a concern about anyone: the sender, a relative, a neighbour or a stranger. This sends the message to a trained person the same hour; it is not a judgement that anyone is at risk. Yes: The message describes signs that someone may be harmed, neglected, unable to look after themselves, or not safe. No: The message raises no such concern.
Does the sender say they are homeless or may soon lose their home? Yes / No
Count a notice to leave, being told to go, sofa-surfing, sleeping rough, or losing a home through eviction or repossession. Yes: The sender says they, or the household they write for, are homeless or may lose their home. No: No such situation is described.
Does the sender express dissatisfaction with a council service, decision or member of staff? Yes / No
Count dissatisfaction anywhere in the message, whether or not the sender uses the word "complaint" and whatever else they ask for. Yes: The sender is unhappy with something the council did, failed to do or decided. No: The message expresses no dissatisfaction with the council.
Does the sender write as, or on behalf of, a councillor or MP? Yes / No
Yes: The message comes from a councillor, an MP or their office, or says it is sent on their behalf. No: The sender writes for themselves or another organisation.
Does the message raise a second matter that belongs to a different service? Yes / No
Count a matter only if a different service from the council service list would have to act on it. Yes: At least two matters in the message belong to different services. No: Everything in the message belongs to one service.
Is information missing that the service needs before it can act? Yes / No
Consider the main matter: where it is, what happened, who it concerns, or which account or case it is about. Do not check whether a reference or address is valid. Yes: The service could not act without asking the sender for more. No: The message gives enough for the service to start.
Does the sender ask for a phone call or to speak to a person? Yes / No
Yes: The sender asks to be phoned, to be called back, or to speak to someone. No: The sender asks for no call and no conversation.
Lens columns
service, service_probability, request_type, request_type_probability, urgency, urgency_average, person_at_risk, person_at_risk_probability, homeless_or_threatened, homeless_or_threatened_probability, complaint, complaint_probability, member_enquiry, member_enquiry_probability, second_matter, second_matter_probability, missing_details, missing_details_probability, callback_requested, callback_requested_probability
This prism on its own pageSend every support ticket to the right team with its urgency
Fields
- Subject
- Ticket text, or the whole thread
Context
We are Pellmore, an online shop that sells our own wireless speakers and headphones, with a companion app and an optional paid subscription, Pellmore Plus, for extra features. Customers write in by email, the web form and in-app chat, and every ticket lands in one helpdesk. Each new ticket is read here and sent to one team. A person in the triage team handles any ticket the answers are unsure about. Routing rules: - One ticket, one team. When a ticket raises more than one issue, route it by the issue that costs the customer most if it waits: money taken wrongly, a device that cannot be used, or a safety concern, in that order of weight. A second issue is split off by code. - Anything about a device getting hot, smoking, swelling, sparking or hurting someone goes to the product safety team first, whatever else the ticket says. - Requests about personal data, and letters from lawyers or authorities, go to privacy and legal. - Urgency is judged from what the customer says about time and impact, not from how serious the topic sounds in general. - The helpdesk matches order numbers, emails and serial numbers to accounts. Nothing here looks them up.
Questions
Which team should handle this support ticket? Choice
Choose the one team that should handle the ticket, using the routing rules in the context. When the ticket raises more than one issue, choose the team for the issue the rules give most weight to.
What does the customer mainly want in this support ticket? Choice
Choose the single main thing the customer wants from the whole ticket, not every thing they mention.
How urgent is this support ticket, from what the customer says? Scale
Rate the time pressure and impact the customer states or clearly describes. Do not rate how important the topic sounds in general.
Does the customer ask to speak to a person, a manager or someone by phone? Yes / No
Read every customer message. Count a request, in any words, for a person, an agent, a manager, a callback or a phone call instead of the assistant, an automated reply or more messages, and a request to stop talking to a bot. Mentioning a person they spoke to before is not a request. Yes: The customer asks, in any words, to deal with a person. No: The customer does not ask for a person, including when they are only frustrated.
Does the team need details from the customer that the ticket does not give, before it can start? Yes / No
Count only details a ticket like this should carry: which order, account email, device or serial number, or what actually went wrong. The helpdesk matching an email address to an account counts as having the account. Judge only the issue the ticket is routed for. Card details, payment IDs and bank statements are not needed; the helpdesk matches charges to the account. Yes: At least one such detail is needed and missing from the whole ticket. No: The ticket carries enough to start work, or the request needs no such detail.
Does the ticket raise a second issue that belongs to a different team? Yes / No
Count a second issue only when the customer asks for something about it and another team in the list would handle it. Yes: The ticket raises a second issue for a different team. No: Everything in the ticket belongs to one team.
Does the customer describe a device or charger getting hot, smoking, swelling, sparking or hurting someone? Yes / No
Count it however it is worded ("the case is bulging", "it smelt of burning"), even when it is not the main point of the ticket. Do not count a device that is only warm in normal use as the customer describes it. Yes: The customer describes a safety concern of this kind. No: No such concern is described.
Lens columns
team, team_probability, intent, intent_probability, urgency, urgency_average, asks_for_person, asks_for_person_probability, missing_details, missing_details_probability, second_issue, second_issue_probability, safety_concern, safety_concern_probability
This prism on its own pageRoute the broker inbox: submissions, binds, changes, claims
Fields
- Email subject
- Email body
- Attachment file names
Context
We are a London MGA and Lloyd's coverholder writing UK commercial property and casualty through brokers. Brokers write to one shared underwriting mailbox. Each email is read here on arrival and sent to one desk: new business, renewals, mid-term changes, the bind desk, documents, broker accounts, or claims, which we pass to our claims administrator. The operations team works anything the answers are unsure about. Rules: - Binding, going on cover and holding a client covered are done by an underwriter, never here. A request for any of them goes to the bind desk first, whatever else the email asks. - New business submissions go on to be read against our appetite by another check. - Claims are handled by our claims administrator under its own authority. A claim, a loss or a circumstance that may lead to a claim is passed to them the same day. - Any dissatisfaction from a broker or a client goes to compliance, who decide whether it is a complaint. - Our systems match quote and policy references. Nothing here looks them up or reads dates.
Questions
What is the broker mainly asking us to do in this email? Choice
Choose the request that needs the fastest action. A request to bind or hold covered always comes first, even under a subject line about something else. Other requests in the same email are recorded by the flags.
Does the broker ask us to bind, go on cover or hold the client covered, anywhere in the email? Yes / No
Count a firm order, "please bind", "please confirm cover from today", or a request to hold covered, even as a side point. Asking for a quote, or saying the client may accept later, does not count. Yes: The email asks us to bind, go on cover or hold covered. No: The email makes no such request.
Does the broker say cover or proof of cover is needed today, or that the client is uninsured until we act? Yes / No
Rely on words such as "today", "by 3pm", "this morning" or "before they can open". Do not compare dates with today; code reads any date. Yes: The broker says cover or proof of cover is needed today or within hours, or that the client has no cover until we act. No: No such urgency is stated.
Does the broker chase something they asked for or sent before? Yes / No
Yes: The broker follows up at least one request, document or answer they sent before. No: Everything in the email is asked for the first time.
Does the email supply information we asked for as a condition of a quote or of cover? Yes / No
Count an attachment when its file name or the email says what it is. Yes: The email sends at least one item we asked for as a condition, such as a certificate, a survey response, a risk improvement confirmation or a signed statement of fact. No: The email supplies nothing of this kind.
Does the email report a claim, a loss or a circumstance that may lead to a claim? Yes / No
Count a report anywhere in the email, even as a side point. A question about claims history on a new submission does not count. Yes: The email reports a new loss, a claim made against the client, or an incident that may lead to one. No: No such report appears.
Does the broker or the client express dissatisfaction with our service, terms or decisions? Yes / No
Count dissatisfaction anywhere in the email, whether or not the word "complaint" is used, and whether the broker speaks for themselves or passes on what the client said. A neutral follow-up is not dissatisfaction. Yes: The broker or the client is unhappy with something we did, failed to do, offered or decided. No: The email expresses no dissatisfaction with us.
How soon must someone act on this broker email? Scale
Rate the most urgent request in the email. Do not compare dates with today; use words such as "today", "by 3pm" or "next week".
Is information missing that the desk needs before it can act? Yes / No
Consider the main request: which client, risk or policy it is about, what cover is wanted, and from when. Do not check whether a reference is valid or read any date; code does that. Yes: The desk could not act without asking the broker for more. No: The email gives enough for the desk to start.
Does the broker ask for a phone call or to speak to a person? Yes / No
Yes: The broker asks to be phoned, to be called back, or to speak to someone. No: The broker asks for no call and no conversation.
Lens columns
request_type, request_type_probability, bind_or_hold_covered, bind_or_hold_covered_probability, cover_needed_urgently, cover_needed_urgently_probability, is_chaser, is_chaser_probability, subjectivities_answered, subjectivities_answered_probability, claim_mentioned, claim_mentioned_probability, complaint_expressed, complaint_expressed_probability, urgency, urgency_average, missing_details, missing_details_probability, callback_requested, callback_requested_probability
This prism on its own pageRoute forwarder emails to the right desk, held cargo first
Fields
- Subject
- Latest message, without the quoted thread
- Attachment file names (empty if none)
Context
We are an air and ocean freight forwarder and customs broker with branches across Europe. Customers (shippers and consignees), carriers, origin and destination agents, and authorities write to each branch's shared operations inbox. Code strips the quoted thread, so the body is the latest message only. Each email is read here on arrival and sent to one desk's queue. A duty coordinator handles anything the answers are unsure about. Emails arrive in several languages; judge them in the language they are written in. Desks: - Quotes and pricing: rate requests for new shipments. - Bookings: new bookings and load offers. - Operations (one per mode): status, amendments, pickup and delivery slots, and partner updates on booked shipments. - Documentation: bills of lading and air waybills, certificates, packing lists and commercial invoices. - Customs: duties, tariff classification, clearance and holds by customs. - Claims: damage, shortage and loss. - Billing: questions about our invoices and charges. - Dangerous goods desk: copied on any mention of dangerous or hazardous goods. A trained specialist checks every one. Routing rules: - One email, one desk. When an email raises more than one matter, route it by the matter that most needs someone to act now. Cargo that has stopped, or a cut-off being missed, comes first. - Nothing here looks up references, container numbers or bookings. Code matches them to the job. - The answers describe the email, never the member of staff who handles it.
Questions
What does the sender mainly want from us in this email? Choice
Pick the matter that most needs someone to act now, using the routing rules in the context. A second matter is recorded by its own question.
Which transport mode does the email concern? Choice
Judge from the words, equipment and references in the email: a 40HC or 20DV container is ocean full container load, an air waybill is air.
How soon must someone act on this email, from what the sender says? Scale
Rate the most urgent matter in the email. Do not compare dates with today; rely on words such as "yesterday", "tonight", "the vessel sails tomorrow" or "still not on board". The word "urgent" in a subject line is not enough on its own.
Does the email refer to a shipment already booked with us? Yes / No
Count a booking or job reference, a container, bill of lading or air waybill number, or plain words about a shipment we are already moving. Yes: The email is about a shipment already booked with us. No: The email is about a new shipment, or about no shipment.
Does the email say cargo is held, selected for inspection or stopped by customs or another authority? Yes / No
Count a hold, an inspection, an examination, a document check or a stop by customs, port health, police or any other authority, at origin or destination, even when the sender is not sure it is true. Yes: The email says cargo is, or may be, held or inspected by an authority. No: The email mentions no hold or inspection by an authority.
Does the email say a cut-off, sailing, flight or delivery slot has been missed or may be missed? Yes / No
Count a cargo, documentation or customs cut-off, a vessel or flight the cargo is not on, or a slot that will be missed. Do not compare dates yourself. Yes: The email says one of these has been missed or is at risk. No: The email says nothing has been or may be missed.
Does the email mention dangerous or hazardous goods? Yes / No
Count a UN number, a hazard class, "DG", "hazardous", "IMO", a safety data sheet, or goods often dangerous in transport such as lithium batteries, aerosols, paint, fuel or gas. This only copies the dangerous goods desk; it never clears anything. Yes: The email mentions dangerous goods or goods often dangerous in transport. No: The email mentions none.
Does the sender express dissatisfaction with our service? Yes / No
Count dissatisfaction in any words: not being told, having to chase, delays we caused, mistakes. Pressure from the sender's own customer counts only if the sender is unhappy with us. Yes: The sender is unhappy with something we did or failed to do. No: The sender expresses no dissatisfaction with us.
Does the email raise a second matter that a different desk would act on? Yes / No
Use the desks in the context. Count a matter only if a different desk from the one the main matter belongs to would have to act on it. Yes: At least two matters in the email belong to different desks. No: Everything in the email belongs to one desk.
Is information missing that the desk needs before it can act? Yes / No
Consider the main matter: which shipment or booking it is about, what cargo, where from and to, and what the sender wants done. Do not check whether a reference or container number is valid. Yes: The desk could not act without asking the sender for more. No: The email gives enough for the desk to start.
Does the sender ask for a phone call or to speak to a person? Yes / No
Yes: The sender asks to be phoned, to be called back, or to speak to someone. No: The sender asks for no call and no conversation.
Lens columns
request_type, request_type_probability, mode, mode_probability, urgency, urgency_average, existing_shipment, existing_shipment_probability, customs_hold, customs_hold_probability, cutoff_missed, cutoff_missed_probability, mentions_dangerous_goods, mentions_dangerous_goods_probability, complaint, complaint_probability, second_matter, second_matter_probability, missing_details, missing_details_probability, callback_requested, callback_requested_probability
This prism on its own pageSee it on your own records
Prismlet is invite only for now. Tell us what you would send it and we will set you up.