Before a CRM or AI, organise the data you already have

If your business sends out quotes, your management software may already know how many were accepted, how many were turned down and how many are still waiting for an answer. What it may not know is why. And buying another tool will not fix that.

This article starts from a simple idea: before thinking about a CRM, automating your follow-up or adding artificial intelligence to your sales, it is worth looking at what data you already have and what state it is in. Quotes are a good example because almost every business sends them and very few make the most of them.

In one sentence: a new tool, whether a CRM or AI, can only work with the data you give it. If the status of your quotes is never updated, or the reason for turning one down sits in free-text notes, sorting that out is the first step and usually the cheapest.

Data weighs almost as much as money

When we ask ourselves why a business does not use AI, we tend to think of cost. Official figures paint a rather different picture.

Among Spanish companies with 10 or more employees that considered using AI but did not go ahead, 48.1% cited problems with the availability or quality of their data. Cost was cited by 47.4%. The two carry practically the same weight.

Source: ONTSI (Spain's National Observatory for Technology and Society), Indicators of Artificial Intelligence Use in Spain, 2026 edition, based on the Spanish National Statistics Institute (INE) survey on ICT use in companies (2025 data). Each company could select more than one reason.

The difference between those two barriers matters. The cost of a tool is not up to you. The state of your data, to a large extent, is. And organising data does not require you to buy anything.

Not an isolated finding

Other reference sources, using different samples and questions, agree in placing data among the main obstacles:

• Across the European Union, Eurostat ranks the availability or quality of data ahead of cost: it was cited by 43.5% of companies with 10 or more employees that had considered using AI, compared with 38.4% citing cost (2025 data).
• The Bank of Spain, using its own business survey, highlights three obstacles among companies that were not using AI and had no plans to do so in the short term: lack of skilled staff (45.8%), cost (40.8%) and the availability or quality of data (37.8%). These are the shares of companies that rated each obstacle as significant or highly significant (data from late 2024).
• The OECD, in its 2025 report for the G7 on AI in small and medium-sized enterprises, lists access to data, algorithms and computing power among the prerequisites for a small business to adopt AI successfully, alongside connectivity, skills and finance.

There is also a study that measures the state of data directly. The researchers Tadhg Nagle, Thomas C. Redman and David Sammon analysed 75 data quality assessments carried out over two years across a range of organisations, data sets and business processes. On average, 47% of recently created records had at least one critical error, and only 3% of the assessments reached an acceptable level (at least 97% of records free of critical errors). It is not a sample of Spanish small businesses, so it should be read as a signal rather than a figure that can be transferred directly. But the authors' conclusion is clear: data tends to be in far worse shape than the people who manage it realise.

What your software may already record

Several management and invoicing programs let you create quotes and mark their status: pending, accepted or rejected. Some go considerably further. One of the dental practice management programs I have analysed, for example, lets you filter quotes by status, dates, person responsible, reason for rejection, financing, communication channel and follow-up actions with their outcome, and then export the list. The documentation of a general invoicing program and of a CRM shows something similar: the first lets you customise quote statuses and the second includes a configurable list of reasons for lost deals.

In other words, in many businesses the sales pipeline already exists inside the software people use every day. What is missing is not a place to store the information, but for the information to be entered properly.

A word of caution before going on: not every program has these fields, and in many businesses quotes live in a word-processing document and in email. If that is your case, the first step is even more basic, and you will find it further down.

Data that exists is not necessarily data you can use

Having the field does not mean having the information. There are three common situations in which a piece of sales data exists but is no use for making any decision.

1. The empty field. The program has a field for the reason a quote was rejected, but nobody fills it in, either because it is not mandatory or because there is never time.

2. Free text. The reason is there, but written differently every time: ‘too expensive’, ‘thinking about it’, ‘payment terms don’t work for them’, ‘needs to talk it over at home’. A person understands it when they read it, but it cannot be counted. And if it cannot be counted, there is no way of knowing whether the most common problem is price, payment terms or a lack of follow-up.

3. The status that never changes. The quote was marked as pending on the day it was sent and nobody has touched it since. Over time, the list of pending quotes mixes live ones with others that nobody is ever going to accept, and the figure stops meaning anything.

None of these three is solved by a new tool. If you connect a CRM or AI to a list like that, it inherits the same mess, only faster.

How to sort it out in a week

You do not need a project. You need to make a few decisions and stick to them.

1. A closed list of rejection reasons. Between five and seven options that cover what really happens in your business, plus an ‘other’ option with a short note. For example: price, payment terms or financing, timing, chose another option, not the right moment, and no reply. You decide the list. What matters is that it is closed.

2. Statuses with dates and a rule for quotes left hanging. Decide how long a quote can stay pending without a reply before it moves to a status of its own, such as ‘no reply’. The time limit depends on your sector. What matters is that the rule exists and is applied.

3. One person responsible for each quote. Not to point the finger at anyone, but so that there is always someone who knows where each quote stands.

4. Follow-up in fields, not in notes. Date of contact, channel and outcome. If your software does not have those fields, a shared spreadsheet is enough to get started.

5. Half an hour a month to look. Export the list and count: how many were accepted, how many were rejected and why, and how many are still awaiting a reply. That half hour is what turns a record into information.

If your quotes live in word-processing documents and email, step zero is a spreadsheet with one row per quote and one column for each of the points above. It is not elegant, but it is structured, and that is what counts.

Where AI comes in, and where it does not

Once the data is in order, AI can genuinely help, and quite a lot.

• Classifying past records. If you have months of free-text notes, a language model can suggest which reason on your closed list each one corresponds to. Its suggestions need checking, because it makes mistakes with ambiguous or very short notes.
• Spotting what has been left hanging. Flagging quotes that have been pending for too long without any recorded follow-up.
• Preparing the follow-up. Drafting a first message tailored to each case for a person to review and decide whether to send.

And three limits worth being clear about:

• AI does not know what nobody has recorded. If the real reason for a rejection is not written down anywhere, it cannot work it out.
• Not all data can be taken to just any tool. At a clinic, a quote describes a treatment, and that is data concerning health, which the EU General Data Protection Regulation (GDPR) treats as a special category (Article 9). Before using an external tool with that data, you need to check that the processing is permitted and, if the provider will process it on the business’s behalf, put a data processing agreement in place with them (Article 28).
• Organising is not hoarding. The same regulation requires personal data to be adequate, relevant, limited to what is necessary and accurate (Article 5). A closed list of reasons helps with both principles: you keep less free text about people, and what you keep is clearer and easier to keep up to date.

When you really do need a CRM

Organising what you already have is not always enough. A CRM makes sense when several people sell and need to share each customer’s history, when the sales cycle is long and there are many contacts before a quote is issued, when you manage opportunities that have not yet reached the quote stage, or when your software does not let you add fields or export your data.

The difference lies in the order of the steps. If you first decide which statuses, which reasons and what follow-up you want to record, any CRM you choose will be set up around those decisions. If you buy it first, you will most likely end up recreating the same mess inside it.

A test for this week

Export your quotes from the last three months and answer three questions:

1. How many are still marked as pending?
2. In how many of the rejected ones has the reason been filled in?
3. Can that reason be counted, or is it free text?

With those three answers you will know whether what you need right now is a new tool or a little order in the one you already have.

Want to know which of your business data is ready for a CRM or AI? Start with the diagnosis. AI Diagnosis €99

Frequently asked questions

Do I need a CRM to follow up on my quotes?

Not necessarily. Several management programs already let you record the status of each quote, and some also record the reason for rejection and follow-up actions. Before buying a CRM, it is worth checking which fields your software has and whether they are filled in consistently.

Why does data quality hold back the use of AI in companies?

Because any AI tool works with the data it receives. According to 2025 INE data compiled by Spain’s ONTSI, 48.1% of companies with 10 or more employees that considered using AI but did not go ahead cited problems with the availability or quality of their data, a weight similar to that of cost (47.4%). Across the European Union, according to Eurostat, the availability or quality of data weighs even more than cost (43.5% compared with 38.4%).

Are data quality problems only an issue for small businesses?

No. A study published in Business Horizons by Nagle, Redman and Sammon, based on 75 assessments across a range of organisations and processes, found that on average 47% of recent records had at least one critical error. The authors link this to a widespread tolerance of living with incorrect data and call on organisations of all sizes to measure the quality of their own data.

What is structured data in a quote?

Data stored in a field with defined values that can be filtered and counted: the status, the date, the person responsible or a rejection reason chosen from a closed list. A free-text note can be read, but not counted.

Can I use AI to classify the rejection reasons I have in notes?

Yes. A language model can suggest which category each note belongs to, but the result should be checked. If the notes contain personal data, and above all data concerning health, you first need to check that the processing is permitted and, if the provider processes it on the business’s behalf, have a data processing agreement in place.

Sources

• ONTSI (Red.es). Indicadores de uso de inteligencia artificial en España. Edición 2026, datos 2025 (in Spanish).
• INE. Encuesta sobre el uso de TIC y del comercio electrónico en las empresas (in Spanish).
• Eurostat. The use of artificial intelligence (AI) technologies in the European Union. Key results, 2026 edition (2025 data).
• Bank of Spain. Fernández Cerezo, A., Hidalgo, I. and Izquierdo, M. Adoption of artificial intelligence in Spanish firms: an initial analysis based on the Banco de España Business Activity Survey. Economic Bulletin 2025/Q2.
• OECD. AI adoption by small and medium-sized enterprises: OECD discussion paper for the G7. December 2025.
• Nagle, T., Redman, T. C. and Sammon, D. Assessing data quality: A managerial call to action. Business Horizons, 63(3), 2020, pp. 325-337.
• Regulation (EU) 2016/679, General Data Protection Regulation, official text, Articles 5, 9 and 28.