Postgres 14

By using db<>fiddle, you agree to license everything you submit by Creative Commons CC0.
CREATE TABLE tbl (id int, company_id int, employee text);
   
INSERT INTO tbl VALUES
  (1, 1, 'Anna')
, (2, 1, 'Bill')
, (3, 2, 'Carol')
, (4, 2, 'Dave')
, (5, 3, 'Tom')
;
5 rows affected

-- simple with GROUP BY:
SELECT company_id, string_agg(employee, ', ' ORDER BY employee) AS employees
FROM   tbl
GROUP  BY company_id
ORDER  BY count(*), company_id;


company_id
employees


3
Tom

1
Anna, Bill

2
Carol, Dave


-- typically faster to sort in a subquery:
SELECT company_id, string_agg(employee, ', ') AS employees
FROM  (SELECT company_id, employee FROM tbl ORDER BY 1, 2) t
GROUP  BY company_id
ORDER  BY count(*), company_id;


company_id
employees


3
Tom

1
Anna, Bill

2
Carol, Dave


-- as academic proof of concept: the same without any aggregate or window functions
WITH RECURSIVE rcte AS (
   (
   SELECT DISTINCT ON (1)
          company_id, employee, ARRAY[employee] AS employees
   FROM   tbl
   ORDER  BY 1, 2
   )
   UNION ALL
   SELECT r.company_id, e.employee, r.employees || e.employee
   FROM   rcte r
   CROSS  JOIN LATERAL (
      SELECT t.employee
      FROM   tbl t
      WHERE  t.company_id = r.company_id
      AND    t.employee   > r.employee
      ORDER  BY t.employee
      LIMIT  1
      ) e
   )
SELECT company_id, array_to_string(employees, ', ') AS employees
FROM  (
   SELECT DISTINCT ON (1)
          company_id, cardinality(employees) AS emp_ct, employees
   FROM   rcte
   ORDER  BY 1, 2 DESC
   ) sub
ORDER  BY emp_ct, company_id;


company_id
employees


3
Tom

1
Anna, Bill

2
Carol, Dave