Commercial outreach at scale has always had a fundamental problem: it is labour-intensive. Find companies, research their situation, personalise the message, send it, follow up. Multiplied by the volume needed for the channel to work, it required a dedicated team of people. And for an SME without that team, the channel was simply not accessible.
Artificial intelligence has changed that. Not in a magical or immediate way, but through a precise system that combines data extraction, automated analysis and the creation of personalised messages. That is the system we use at BAI, and what we have built as a product for our clients.
The problem with generic outreach
Before explaining how it’s done, we need to understand why generic outreach doesn’t work.
The inbox of any company director is full of emails following this template: "Hi [Name], we are [Company] and we offer [Generic Service]. Could we have a fifteen-minute call?" The response rate for this type of message is between zero and one per cent. And that rate doesn’t improve by sending more: only by sending better.
What truly works is a message that shows the sender has studied the recipient’s business in detail, identified a specific problem, and is proposing something meaningful for that problem. Genuine personalisation generates response rates five to fifteen times higher than generic outreach.
Until now, the problem was that genuine personalisation required human time. No longer.
The system in five steps
1. Step: Data extraction. The system begins identifying companies that match the target profile. In the case of a digital marketing agency, the profile covers companies with an outdated website, no presence on Google Maps, no activity on social media, and operating in sectors that have historically responded well to digital marketing services.
The data sources are: OpenStreetMap for creating lists of local businesses, Páginas Amarillas and sector directories for contact details, and Google Maps for information on digital presence. For each company found, the system extracts the name, sector, municipality, website, email address and phone number when available.
Step 2: Digital presence analysis. For each company with a website, the system analyses it automatically. PageSpeed API for load speed and Core Web Vitals. Wayback Machine for the age and frequency of updates. Monitoring of social media presence and Google Business profile status. The result of this analysis is a score from 0 to 100, indicating its potential as a prospective client.
Step 3: Filtering and prioritisation. Companies with a high score have more digital problems to solve and more sales opportunities. These are the ones that receive outreach first. Companies with a low score — an up-to-date website, good SEO, an active presence — have no urgent need and are less likely to respond positively.
4. Step: Creation of a personalised message. This is where the language model comes in. For each company that passes the filter, the system generates a specific email that mentions the exact problems detected in the analysis: "I visited your website and saw that it takes 8 seconds to load on mobile, that your Google Maps profile is unclaimed, and that the last website update dates from 2021. This directly affects how many customers find you when searching on Google."
This message is not a template with replaced variables. It is text created for this specific company, containing its specific data, and it shows that someone has analysed its business.
5. Step: Controlled sending and response management. Sending is deliberately controlled: between fifty and eighty emails per day. That limit is not arbitrary — it is the range that allows the sending domain’s reputation and high delivery rates to be maintained. Emails that do not reach the inbox generate no responses.
Responses are detected automatically. Leads that respond with interest are passed to the CRM, with the full context of the previous analysis and the conversation history. The sales team only intervenes at that point.
The numbers you can expect
When this system is running at full capacity, you can contact between one thousand and one thousand five hundred companies per month. Response rates of between three and five per cent —realistic for well-personalised outreach— generate between thirty and seventy-five sales conversations per month. Depending on the conversion rate of your sales process, this can mean between five and twenty new clients per month.
For a company that has never had an active prospecting process, this transforms the acquisition channel. Without hiring a sales team, without paying for advertising campaigns per lead, with a monthly operating cost much lower than the alternatives.
What this system does not replace
The system generates conversations. It doesn’t close them. Commercial proposals, negotiation, customer relationship management: these still require human judgement.
Furthermore, you also need a product or service that is genuinely relevant to the problems the analysis identifies. If outreach identifies that a company has a slow website, but your digital marketing services do not include web optimisation, the match does not work.
And it requires a fast-response process. Leads who arrive through well-personalised outreach have a window of interest that can last hours or days. If you need a week to reply, that interest will have gone cold.