AI on the admin around selling, not on the selling

AED 20,000 to AED 120,000 covers most of what we are asked to build. The number moves on how many workflows, whether CRM integration needs building, how many users, and whether ongoing accuracy review is included. Tool and model usage costs are separate and ongoing. Anything quoted before those are known is a guess. We scope first, then price, and the scope document is yours whether or not you proceed. These are market ranges rather than a fixed rate card — the proposal carries the real figure.

No. Automated outreach at volume, however well personalised it appears, is recognisable to recipients, has collapsing reply rates as adoption has spread, and puts your sending domain at risk. We will build everything around the conversation and leave the conversation with a person. If that is the specific thing you want, there are vendors who will sell it and we are not one.

Call recording, transcription and summarisation with automatic CRM population. It is the highest-return, lowest-risk item on the list, reps like it immediately because it removes real hours, and it improves data quality as a side effect because the alternative was usually an empty field.

First drafts from your templates and the deal record, yes, for a person to finish. Sending an unreviewed generated proposal to a client is a commercial risk rather than an efficiency, and the review step is where the accountability sits.

Pattern-matching on your own pipeline history, yes — deals that have gone quiet, stages held longer than normal, missing committee contacts. It is useful as a prompt for a human to look, not as a verdict, and it needs enough closed history to mean anything.

Consent rules matter and they vary by jurisdiction and by who is on the call. We check the position for your situation before building anything that records, and it is the first thing to resolve rather than the last. Notifying participants is the safe default regardless.

That depends entirely on which tools are chosen, and it is a decision rather than an accident. We select tooling with appropriate data handling, avoid consumer-tier products for anything containing client data, and write the rules down. If this matters to you, ask every vendor directly and get it in writing.

Spot-checking on a schedule, plus a review step before summaries write to the CRM. In the first weeks reps confirm every summary, which both catches errors and builds the trust that determines whether they keep using it.

Four to ten weeks depending on how many workflows and how much CRM integration is needed. Starting with call summarisation alone can be live in under a month.

It depends on how much admin your reps currently do, which is why we measure it first. The recurring pattern is that note-taking and CRM updating consume more hours than management estimates, and that is where the bulk of the saving comes from. We report hours against the baseline rather than quoting a benchmark from a vendor study.

Yes, and we would check that first. Several CRM and conference platforms now include transcription and summarisation, and if you are already paying for it, using it beats buying another tool.

Enrichment and summarisation, yes, and it is a reliable win. Firmographic data filled in, recent news summarised, previous conversations pulled together into a brief before a call. What it should not do is decide which leads are worth pursuing on its own, because the reasoning is not inspectable and a rep who cannot see why will not trust it.

It works better than most people expect for the mechanical parts — talk ratio, how long before a question was asked, whether the next step was agreed. It is much weaker at judging whether the conversation was any good, and a manager listening to two calls a week still beats it. Use it to surface which calls a manager should listen to.

Not on its own, and treat any claim otherwise carefully. What it can do is flag deals whose behaviour resembles ones that historically slipped, which is a useful prompt for a human to look. A forecast a sales director cannot interrogate does not get used, regardless of accuracy.

Some, and less than you might fear. Call summarisation and transcription can run on a messy CRM and will actually improve it, because populated fields beat empty ones. Anything predictive needs clean history, which is why we sequence it that way.

Considerably less than with most CRM changes, in our experience, because the first thing it does is remove work rather than add it. The resistance appears when the automation starts monitoring them rather than helping them, which is a scoping decision you make rather than an inevitability.