Grading leads A to D with AI
How Achekam's AI grades a lead A to D against a weighted rubric, what happens without AI, and when staff should override a criterion.
By Achekam · · 5 min read · For Accountants & audit firms in South Africa, Nigeria, Kenya
Monday morning, and the firm's lead list has grown over the weekend: a few enquiries from the website, a batch from a Facebook campaign, a referral from an existing client. Nobody has time to call all of them before lunch. Somebody has to decide which ones first, and in most firms that decision is made by instinct, by whoever happens to open the list.
Grading tries to make the first pass more consistent. Achekam can have AI grade each lead from A to D, so the people who follow up start from a common view of which leads look most promising. The value of that grade depends entirely on understanding what it is: advice that a person can correct, not a decision.
The weighted rubric
From a lead's page, staff can ask AI to grade the lead A to D against a weighted rubric. The rubric is a set of criteria, each carrying a weight, and the grade summarises how the lead scores across them. The grade is saved on the lead for the whole firm, so the partner who looks at the list in the afternoon sees the same grade the administrator saw in the morning.
Consistency is the main benefit. Two people looking at the same thin enquiry, a name, an e-mail address and a sentence about "help with tax", will often rank it differently. A rubric applied the same way to every lead gives the team a shared starting point.
What AI cannot do is know what the enquiry form did not capture. A lead graded D on its form answers may be the brother-in-law of the firm's biggest client. That is why the rubric can be overridden.
Fixed rules when AI is not available
Achekam's AI features go through one AI engine that can use Anthropic, OpenAI, Google Gemini, DeepSeek or xAI, using the firm's own API key first and the platform's key only where allowed. With no key at all, every AI feature says so and drafts nothing.
Grading does not simply stop when AI is unavailable. It falls back to fixed rules, and the lead is still graded A to D. A firm should know which method produced a given grade, because a rules-based grade and an AI grade are built differently, and staff should not treat them as interchangeable when comparing leads graded on different days.
Overriding a criterion
Staff can override any criterion in the rubric. That is the point at which the person's knowledge enters the grade.
Good reasons to override:
- Information from a conversation. After a first call, the lead's size, urgency or budget is clearer than the form suggested.
- A relationship the form cannot show. A referral from a valued client, or a former client returning.
- A service the firm does not offer. A lead asking only for work the firm declines should not sit at the top of the list.
- A risk the firm has spotted. An obvious conflict with an existing client, or a request the firm would not accept.
Poor reasons to override are just as easy to recognise: liking the sound of the person's name, or moving a lead up because it is assigned to you. Agree as a team that every override has a reason someone else could understand, and write that reason where colleagues will see it.
The grade is advice, not a decision
Nothing acts on the grade automatically. It does not assign the lead, enrol it in a sequence or close it. Where the firm has switched on automatic assignment of unowned connector leads by turn or by audience, or a first-response target, those work on their own settings, not on the grade; see assigning new leads and response targets.
This reflects a wider rule in the AI engine itself: it refuses any task whose type is giving an opinion, reaching a conclusion, approving or signing. A grade helps a person decide whom to call first. A person still decides.
Treat it in team routines accordingly:
| Grade | A sensible default |
|---|---|
| A | Call today; owner checks the grade after the first conversation |
| B | Contact within the firm's normal response time |
| C | Follow up by e-mail; review the grade if the lead replies |
| D | Check whether it is genuinely a poor fit before closing it |
The last row matters. A D grade is a reason to look, not a reason to discard.
What the AI sees, and South African wording
Grading sends the lead's details to an AI vendor. By default, Achekam's AI engine removes South African ID numbers before text is sent, and an administrator can switch that removal off in AI settings. The AI Activity Log lists every AI request with vendor, model, tokens and estimated cost, and a workspace administrator can switch AI off for the workspace.
Sending personal details to a vendor, often outside the country, is a data protection question as well as a technical one. For Nigerian firms, the Nigeria Data Protection Commission's FAQs explain what the Nigeria Data Protection Act 2023 covers and who it applies to, including consent and cross-border transfers. South African and Kenyan firms should apply their own countries' rules in the same way.
One limit applies to Nigerian and Kenyan firms specifically: the grading prompt mentions South African practice even for Nigerian or Kenyan firms. A grade for a Lagos or Nairobi lead may therefore lean on assumptions that fit South Africa better. Staff in those firms should expect to override more often, especially on criteria that depend on local rules or local market conditions.
Leads from social media advertising are where grading earns its keep, because they arrive in volume with little detail; see Facebook lead ads into the firm's CRM via Zapier.
For the next fortnight, ask whoever makes first contact to note whether the grade matched what they found on the call. The pattern of overrides will tell the firm far more about how to use grading than any setting will.
General guidance only, not advice on your own circumstances. Tax and payroll rules change: check the current position with the authority or your adviser.
