Short answer
SME owners and business leaders in larger enterprises can first validate an internal list with sources, named owners and next steps. Check fresh replies and synchronisation the next morning, then have authorised colleagues approve external content; do not automatically send messages, change prices or promise delivery.
An owner may see how many quotations have been sent without seeing which opportunities are waiting for the company’s next move. A customer’s delivery question going unanswered, a sales colleague assuming another department is handling it, or an order status not yet synchronised can all send the next follow-up in the wrong direction.
SME owners need to know: which issues need my decision, and who is responsible for following up? Owners, sales leaders and operations leads in larger enterprises also need to check for gaps in cross-team handovers and identify who approves exceptions. The output worth validating first is a list with sources, named owners and next steps.
To assess what AI could do for your company, start with one question: ‘Which tasks can be done after colleagues finish work, without needing immediate customer contact?’ Organising information before quotation follow-up is a candidate worth examining. Night-time is only a lens for observing the work; whether it can be automated still depends on data, rules, permissions and the consequences of errors.
Define a list that colleagues can actually use
If the system merely lists ‘all quotations that have not converted’, colleagues still have to investigate again. A more useful output includes the quotation number, owner, date of the last meaningful interaction, links to original records, outstanding tasks and a suggested next step. Every judgement must distinguish ‘supported by records’ from ‘needs confirmation’.
For example, when a customer explicitly writes ‘waiting for procurement approval’, that can be recorded as their stated status. If they simply have not replied, the only valid label is ‘no reply found’. AI should not decide that the price is too high, and certainly should not add a discount automatically.
How an after-hours run could work
1. Read authorised CRM data, quotation records and related emails, and record the data cut-off for this run. First exclude test quotations, cancelled items, completed sales and customers who have requested a pause in contact.
2. Match by quotation number and customer identifier. If one customer has several quotations, do not merge them using the company name alone. Put uncertain matches into a verification list.
3. Use explicit rules to select items requiring review, such as those beyond the company’s chosen follow-up interval or those with unanswered customer questions. The business owner sets the interval; there is no universal standard to apply.
4. Have AI organise conversation summaries, confirmed requirements, missing information and draft replies. Attach the original source to every summary; write ‘unconfirmed’ where supporting evidence cannot be found.
5. The next morning, the responsible colleague checks the latest replies before deciding the contact time, content and channel. Quotation amounts, delivery dates, discounts and commitments still require approval by an authorised person.
If new orders arrive overnight but have not yet synchronised, the list must state the synchronisation status so that customers who have already ordered are not chased again in the morning. Initially, the workflow can read data and generate internal lists only, without directly changing customer statuses or sending messages.
Complete synthetic fictional demonstration: turning source records into next-day tasks
All identifiers, dates, roles and quantities below are synthetic hypothetical examples, not real client cases, measured results or revenue forecasts. The demonstration cut-off is 2026-10-14 18:00; source codes serve only to illustrate traceability.
Before organisation: three data sources, three outstanding quotations
CRM test row C01: Q-DEMO-101; owner “Sales A”; status “quotation sent”; last updated 14 October at 16:00.
Email test row E01: linked to Q-DEMO-101; at 15:30 on 14 October the customer asked, “Can delivery be split into two batches?” No reply appears in the test records.
CRM test row C02: Q-DEMO-102; owner “Sales B”; status “awaiting follow-up”. Email test row E02: the customer stated, “We will reply after the procurement meeting on 20 October.”
CRM test row C03: Q-DEMO-103; owner “Sales A”; status “awaiting follow-up”. Order test row O01: an order has been created against the same quotation number, but CRM synchronisation is not yet complete.
After organisation: the next-day internal review list
1. Q-DEMO-101: the known unanswered question concerns split delivery. Sales A should confirm feasible arrangements with operations before drafting a reply; sources C01 and E01. Delivery dates and split-delivery charges are both unconfirmed, so no commitment can be made.
2. Q-DEMO-102: the customer has stated their decision timing. Sales B should arrange follow-up according to the situation after the 20 October meeting; sources C02 and E02. Do not chase an immediate reply simply because the quotation has not converted.
3. Q-DEMO-103: CRM and order statuses conflict. Keep it out of the ordinary reminder list and ask Sales A to check O01 and the latest order; sources C03 and O01. Do not automatically change CRM before verification.
An owner or manager can see at a glance: one item awaits internal information, one awaits the customer’s stated decision timing, and one needs a status check first. The demonstration shows an organisation format; it does not establish how much time was saved or how much business was generated.
Where human control must remain
The next day, check the latest replies and synchronisation status before authorised colleagues approve external content. If a quotation number cannot be matched, sources are missing or data is incomplete, mark the item for verification. The pilot generates internal drafts only; it does not automatically send messages, change prices or promise delivery.
How to decide whether to continue
Before the pilot, record how long each preparation takes, how many systems must be opened and how many items needing follow-up are missed. During the pilot, measure review time, incorrect matches, stale data and unsupported summaries. Subtract review, correction and maintenance time from any preparation time saved, and include tool and integration costs.
A more complete list can support timely handling of opportunities by sales colleagues. Sales and revenue still depend on price, demand and service capacity; any growth should not be attributed directly to AI.
Set boundaries for customer-data use first
Read only the fields needed for the task, restrict access by role, set retention periods and confirm the supplier’s data-use and storage arrangements. Do not paste complete customer lists into public tools that the company has not approved. A pilot can start with de-identified samples, moving to actual records only after the data flow and permissions are established.
Write down one workflow to decide which step merits investment
On a sheet of paper or in a document, write down one task where follow-up is easily missed, where its data is held, who is responsible, the current manual time and the consequences of errors. An SME owner can bring one example; a larger enterprise can start with one business unit and involve the process owner and relevant system administrators.
Write down one workflow to decide which step merits investment
Book a 60-minute AI transformation assessment to review existing systems, time-consuming steps and a suitable scope to validate first. Start by discussing quotation-data organisation, review costs and approval boundaries; no raw customer data is needed for the initial enquiry.
Book a 60-minute AI transformation assessment →Source: Smark Global editorial process design; all scenarios are fictional illustrations, not client cases or evidence of results.
Last updated: 2026-10-14（first published: 2026-10-14）
